Merge branch 'main' into dev/dotnet_workflow/state_manager_object_check

fix: Re-enable the "retrieve object" check in StateManager
.NET: fix: HandoffAgentExecutor does not output any response when non-streaming (#4745 )
2026-06-16 21:04:09 +08:00 · 2026-03-25 18:40:06 -04:00 · 2026-03-25 18:23:52 -04:00 · 2026-03-25 22:22:13 +00:00 · 2026-03-25 20:35:17 +00:00 · 2026-03-25 20:02:16 +00:00
562 changed files with 15312 additions and 12864 deletions
@@ -0,0 +1,166 @@
+name: Setup Local MCP Server
+description: Start and validate a local streamable HTTP MCP server for integration tests
+
+inputs:
+  fallback_url:
+    description: Existing LOCAL_MCP_URL value to keep as a fallback if local startup fails
+    required: false
+    default: ''
+  host:
+    description: Host interface to bind the local MCP server
+    required: false
+    default: '127.0.0.1'
+  port:
+    description: Port to bind the local MCP server
+    required: false
+    default: '8011'
+  mount_path:
+    description: Mount path for the local streamable HTTP MCP endpoint
+    required: false
+    default: '/mcp'
+
+outputs:
+  effective_url:
+    description: Local MCP URL when startup succeeds, otherwise the provided fallback URL
+    value: ${{ steps.start.outputs.effective_url }}
+  local_url:
+    description: URL of the local MCP server
+    value: ${{ steps.start.outputs.local_url }}
+  started:
+    description: Whether the local MCP server started and passed validation
+    value: ${{ steps.start.outputs.started }}
+  pid:
+    description: PID of the local MCP server process when startup succeeded
+    value: ${{ steps.start.outputs.pid }}
+
+runs:
+  using: composite
+  steps:
+    - name: Start and validate local MCP server
+      id: start
+      shell: bash
+      run: |
+        set -euo pipefail
+
+        host="${{ inputs.host }}"
+        port="${{ inputs.port }}"
+        mount_path="${{ inputs.mount_path }}"
+        fallback_url="${{ inputs.fallback_url }}"
+
+        if [[ ! "$mount_path" =~ ^/ ]]; then
+          mount_path="/$mount_path"
+        fi
+
+        local_url="http://${host}:${port}${mount_path}"
+        health_url="http://${host}:${port}/healthz"
+        log_file="$RUNNER_TEMP/local-mcp-server.log"
+        pid_file="$RUNNER_TEMP/local-mcp-server.pid"
+        rm -f "$log_file" "$pid_file"
+
+        server_pid="$(
+          python3 - "$GITHUB_WORKSPACE/python" "$log_file" "$host" "$port" "$mount_path" <<'PY'
+        from __future__ import annotations
+
+        import subprocess
+        import sys
+
+        workspace, log_file, host, port, mount_path = sys.argv[1:]
+
+        with open(log_file, "w", encoding="utf-8") as log:
+            process = subprocess.Popen(
+                [
+                    "uv",
+                    "run",
+                    "python",
+                    "scripts/local_mcp_streamable_http_server.py",
+                    "--host",
+                    host,
+                    "--port",
+                    port,
+                    "--mount-path",
+                    mount_path,
+                ],
+                cwd=workspace,
+                stdout=log,
+                stderr=subprocess.STDOUT,
+                start_new_session=True,
+            )
+
+        print(process.pid)
+        PY
+        )"
+        echo "$server_pid" > "$pid_file"
+
+        started=false
+        for _ in $(seq 1 30); do
+          if curl --silent --fail "$health_url" >/dev/null; then
+            started=true
+            break
+          fi
+          if ! kill -0 "$server_pid" 2>/dev/null; then
+            break
+          fi
+          sleep 1
+        done
+
+        if [[ "$started" == "true" ]]; then
+          if ! (
+            cd "$GITHUB_WORKSPACE/python"
+            LOCAL_MCP_URL="$local_url" uv run python - <<'PY'
+        from __future__ import annotations
+
+        import asyncio
+        import os
+
+        from agent_framework import Content, MCPStreamableHTTPTool
+
+
+        def result_to_text(result: str | list[Content]) -> str:
+            if isinstance(result, str):
+                return result
+            return "\n".join(content.text for content in result if content.type == "text" and content.text)
+
+
+        async def main() -> None:
+            tool = MCPStreamableHTTPTool(
+                name="local_ci_mcp",
+                url=os.environ["LOCAL_MCP_URL"],
+                approval_mode="never_require",
+            )
+
+            async with tool:
+                assert tool.functions, "Local MCP server did not expose any tools."
+                result = result_to_text(await tool.functions[0].invoke(query="What is Agent Framework?"))
+                assert result, "Local MCP server returned an empty response."
+
+
+        asyncio.run(main())
+        PY
+          ); then
+            started=false
+          fi
+        fi
+
+        effective_url="$local_url"
+        pid="$server_pid"
+
+        if [[ "$started" != "true" ]]; then
+          effective_url="$fallback_url"
+          pid=""
+          if kill -0 "$server_pid" 2>/dev/null; then
+            kill -TERM -- "-$server_pid" 2>/dev/null || kill -TERM "$server_pid" || true
+            sleep 1
+            kill -KILL -- "-$server_pid" 2>/dev/null || kill -KILL "$server_pid" || true
+          fi
+          echo "Local MCP server was unavailable; continuing with fallback LOCAL_MCP_URL."
+          if [[ -f "$log_file" ]]; then
+            tail -n 100 "$log_file" || true
+          fi
+        else
+          echo "Using local MCP server at $local_url"
+        fi
+
+        echo "started=$started" >> "$GITHUB_OUTPUT"
+        echo "local_url=$local_url" >> "$GITHUB_OUTPUT"
+        echo "effective_url=$effective_url" >> "$GITHUB_OUTPUT"
+        echo "pid=$pid" >> "$GITHUB_OUTPUT"
@@ -41,8 +41,7 @@ ENFORCED_TARGETS: set[str] = {
    "packages.purview.agent_framework_purview",
    "packages.anthropic.agent_framework_anthropic",
    "packages.azure-ai-search.agent_framework_azure_ai_search",
-    "packages.core.agent_framework.azure",
-    "packages.core.agent_framework.openai",
+    "packages.openai.agent_framework_openai",
    # Individual files (if you want to enforce specific files instead of whole packages)
    "packages/core/agent_framework/observability.py",
    # Add more targets here as coverage improves
@@ -63,6 +63,8 @@ jobs:
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
      OPENAI_EMBEDDINGS_MODEL_ID: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_EMBEDDING_MODEL: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
    defaults:
      run:
@@ -81,8 +83,8 @@ jobs:
      - name: Test with pytest (OpenAI integration)
        run: >
          uv run pytest --import-mode=importlib
-          packages/core/tests/openai
-          -m integration
+          packages/openai/tests
+          -m "integration and not azure"
          -n logical --dist worksteal
          --timeout=120 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5
@@ -94,8 +96,9 @@ jobs:
    environment: integration
    timeout-minutes: 60
    env:
-      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__CHATDEPLOYMENTNAME }}
+      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
      AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      AZURE_OPENAI_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
      AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__EMBEDDINGDEPLOYMENTNAME }}
      AZURE_OPENAI_ENDPOINT: ${{ vars.AZUREOPENAI__ENDPOINT }}
    defaults:
@@ -121,7 +124,9 @@ jobs:
      - name: Test with pytest (Azure OpenAI integration)
        run: >
          uv run pytest --import-mode=importlib
-          packages/core/tests/azure
+          packages/openai/tests/openai/test_openai_chat_completion_client_azure.py
+          packages/openai/tests/openai/test_openai_chat_client_azure.py
+          packages/azure-ai/tests/azure_openai
          -m integration
          -n logical --dist worksteal
          --timeout=120 --session-timeout=900 --timeout_method thread
@@ -151,6 +156,13 @@ jobs:
        with:
          python-version: ${{ env.UV_PYTHON }}
          os: ${{ runner.os }}
+      - name: Start local MCP server
+        id: local-mcp
+        uses: ./.github/actions/setup-local-mcp-server
+        with:
+          fallback_url: ${{ env.LOCAL_MCP_URL }}
+      - name: Prefer local MCP URL when available
+        run: echo "LOCAL_MCP_URL=${{ steps.local-mcp.outputs.effective_url }}" >> "$GITHUB_ENV"
      - name: Test with pytest (Anthropic, Ollama, MCP integration)
        run: >
          uv run pytest --import-mode=importlib
@@ -161,6 +173,26 @@ jobs:
          -n logical --dist worksteal
          --timeout=120 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5
+      - name: Stop local MCP server
+        if: always()
+        shell: bash
+        run: |
+          set -euo pipefail
+          server_pid="${{ steps.local-mcp.outputs.pid }}"
+          if [[ -z "$server_pid" ]]; then
+            exit 0
+          fi
+          if ! kill -0 "$server_pid" 2>/dev/null; then
+            exit 0
+          fi
+          kill -TERM -- "-$server_pid" 2>/dev/null || kill -TERM "$server_pid" 2>/dev/null || true
+          for _ in $(seq 1 10); do
+            if ! kill -0 "$server_pid" 2>/dev/null; then
+              exit 0
+            fi
+            sleep 1
+          done
+          kill -KILL -- "-$server_pid" 2>/dev/null || kill -KILL "$server_pid" 2>/dev/null || true

  # Azure Functions + Durable Task integration tests
  python-tests-functions:
@@ -172,10 +204,13 @@ jobs:
      UV_PYTHON: "3.11"
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
-      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__CHATDEPLOYMENTNAME }}
-      AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      OPENAI_EMBEDDING_MODEL: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
      AZURE_OPENAI_ENDPOINT: ${{ vars.AZUREOPENAI__ENDPOINT }}
+      AZURE_OPENAI_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      FOUNDRY_MODEL: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
+      FOUNDRY_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
      FUNCTIONS_WORKER_RUNTIME: "python"
      DURABLE_TASK_SCHEDULER_CONNECTION_STRING: "Endpoint=http://localhost:8080;TaskHub=default;Authentication=None"
      AzureWebJobsStorage: "UseDevelopmentStorage=true"
@@ -209,7 +244,8 @@ jobs:
          packages/durabletask/tests/integration_tests
          -m integration
          -n logical --dist worksteal
-          --timeout=120 --session-timeout=900 --timeout_method thread
+          -x
+          --timeout=360 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5

  # Azure AI integration tests
@@ -221,6 +257,8 @@ jobs:
    env:
      AZURE_AI_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
      AZURE_AI_MODEL_DEPLOYMENT_NAME: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
+      FOUNDRY_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
+      FOUNDRY_MODEL: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
      LOCAL_MCP_URL: ${{ vars.LOCAL_MCP__URL }}
    defaults:
      run:
@@ -244,7 +282,9 @@ jobs:
          subscription-id: ${{ secrets.AZURE_SUBSCRIPTION_ID }}
      - name: Test with pytest
        timeout-minutes: 15
-        run: uv run --directory packages/azure-ai poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5
+        run: |
+          uv run --directory packages/azure-ai poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5
+          uv run --directory packages/foundry poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5

  # Azure Cosmos integration tests
  python-tests-cosmos:
@@ -47,6 +47,9 @@ jobs:
          filters: |
            python:
              - 'python/**'
+              - '.github/actions/setup-local-mcp-server/**'
+              - '.github/workflows/python-merge-tests.yml'
+              - '.github/workflows/python-integration-tests.yml'
            core:
              - 'python/packages/core/agent_framework/_*.py'
              - 'python/packages/core/agent_framework/_workflows/**'
@@ -54,20 +57,30 @@ jobs:
              - 'python/packages/core/agent_framework/observability.py'
            openai:
              - 'python/packages/core/agent_framework/openai/**'
-              - 'python/packages/core/tests/openai/**'
+              - 'python/packages/openai/**'
+              - 'python/samples/**/providers/openai/**'
            azure:
+              - 'python/packages/openai/**'
              - 'python/packages/core/agent_framework/azure/**'
-              - 'python/packages/core/tests/azure/**'
+              - 'python/packages/azure-ai/agent_framework_azure_ai/_deprecated_azure_openai.py'
+              - 'python/packages/azure-ai/tests/azure_openai/**'
+              - 'python/samples/**/providers/azure/openai_chat_completion_client_azure*.py'
            misc:
              - 'python/packages/anthropic/**'
              - 'python/packages/ollama/**'
              - 'python/packages/core/agent_framework/_mcp.py'
              - 'python/packages/core/tests/core/test_mcp.py'
+              - 'python/scripts/local_mcp_streamable_http_server.py'
+              - '.github/actions/setup-local-mcp-server/**'
+              - '.github/workflows/python-merge-tests.yml'
+              - '.github/workflows/python-integration-tests.yml'
            functions:
              - 'python/packages/azurefunctions/**'
              - 'python/packages/durabletask/**'
            azure-ai:
              - 'python/packages/azure-ai/**'
+              - 'python/packages/foundry/**'
+              - 'python/samples/**/providers/foundry/**'
            cosmos:
              - 'python/packages/azure-cosmos/**'
      # run only if 'python' files were changed
@@ -131,6 +144,8 @@ jobs:
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
      OPENAI_EMBEDDINGS_MODEL_ID: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_EMBEDDING_MODEL: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
    defaults:
      run:
@@ -146,8 +161,8 @@ jobs:
      - name: Test with pytest (OpenAI integration)
        run: >
          uv run pytest --import-mode=importlib
-          packages/core/tests/openai
-          -m integration
+          packages/openai/tests
+          -m "integration and not azure"
          -n logical --dist worksteal
          --timeout=120 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5
@@ -180,8 +195,9 @@ jobs:
    runs-on: ubuntu-latest
    environment: integration
    env:
-      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__CHATDEPLOYMENTNAME }}
+      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
      AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      AZURE_OPENAI_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
      AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__EMBEDDINGDEPLOYMENTNAME }}
      AZURE_OPENAI_ENDPOINT: ${{ vars.AZUREOPENAI__ENDPOINT }}
    defaults:
@@ -205,7 +221,9 @@ jobs:
      - name: Test with pytest (Azure OpenAI integration)
        run: >
          uv run pytest --import-mode=importlib
-          packages/core/tests/azure
+          packages/openai/tests/openai/test_openai_chat_completion_client_azure.py
+          packages/openai/tests/openai/test_openai_chat_client_azure.py
+          packages/azure-ai/tests/azure_openai
          -m integration
          -n logical --dist worksteal
          --timeout=120 --session-timeout=900 --timeout_method thread
@@ -253,6 +271,13 @@ jobs:
        with:
          python-version: ${{ env.UV_PYTHON }}
          os: ${{ runner.os }}
+      - name: Start local MCP server
+        id: local-mcp
+        uses: ./.github/actions/setup-local-mcp-server
+        with:
+          fallback_url: ${{ env.LOCAL_MCP_URL }}
+      - name: Prefer local MCP URL when available
+        run: echo "LOCAL_MCP_URL=${{ steps.local-mcp.outputs.effective_url }}" >> "$GITHUB_ENV"
      - name: Test with pytest (Anthropic, Ollama, MCP integration)
        run: >
          uv run pytest --import-mode=importlib
@@ -264,6 +289,26 @@ jobs:
          --timeout=120 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5
        working-directory: ./python
+      - name: Stop local MCP server
+        if: always()
+        shell: bash
+        run: |
+          set -euo pipefail
+          server_pid="${{ steps.local-mcp.outputs.pid }}"
+          if [[ -z "$server_pid" ]]; then
+            exit 0
+          fi
+          if ! kill -0 "$server_pid" 2>/dev/null; then
+            exit 0
+          fi
+          kill -TERM -- "-$server_pid" 2>/dev/null || kill -TERM "$server_pid" 2>/dev/null || true
+          for _ in $(seq 1 10); do
+            if ! kill -0 "$server_pid" 2>/dev/null; then
+              exit 0
+            fi
+            sleep 1
+          done
+          kill -KILL -- "-$server_pid" 2>/dev/null || kill -KILL "$server_pid" 2>/dev/null || true
      - name: Surface failing tests
        if: always()
        uses: pmeier/pytest-results-action@v0.7.2
@@ -290,10 +335,13 @@ jobs:
      UV_PYTHON: "3.11"
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
-      AZURE_OPENAI_CHAT_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__CHATDEPLOYMENTNAME }}
-      AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      OPENAI_EMBEDDING_MODEL: ${{ vars.OPENAI_EMBEDDING_MODEL_ID }}
      AZURE_OPENAI_ENDPOINT: ${{ vars.AZUREOPENAI__ENDPOINT }}
+      AZURE_OPENAI_DEPLOYMENT_NAME: ${{ vars.AZUREOPENAI__RESPONSESDEPLOYMENTNAME }}
+      FOUNDRY_MODEL: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
+      FOUNDRY_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
      FUNCTIONS_WORKER_RUNTIME: "python"
      DURABLE_TASK_SCHEDULER_CONNECTION_STRING: "Endpoint=http://localhost:8080;TaskHub=default;Authentication=None"
      AzureWebJobsStorage: "UseDevelopmentStorage=true"
@@ -325,7 +373,8 @@ jobs:
          packages/durabletask/tests/integration_tests
          -m integration
          -n logical --dist worksteal
-          --timeout=120 --session-timeout=900 --timeout_method thread
+          -x
+          --timeout=360 --session-timeout=900 --timeout_method thread
          --retries 2 --retry-delay 5
        working-directory: ./python
      - name: Surface failing tests
@@ -352,6 +401,8 @@ jobs:
    env:
      AZURE_AI_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
      AZURE_AI_MODEL_DEPLOYMENT_NAME: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
+      FOUNDRY_PROJECT_ENDPOINT: ${{ secrets.AZUREAI__ENDPOINT }}
+      FOUNDRY_MODEL: ${{ vars.AZUREAI__DEPLOYMENTNAME }}
      LOCAL_MCP_URL: ${{ vars.LOCAL_MCP__URL }}
    defaults:
      run:
@@ -373,7 +424,9 @@ jobs:
          subscription-id: ${{ secrets.AZURE_SUBSCRIPTION_ID }}
      - name: Test with pytest
        timeout-minutes: 15
-        run: uv run --directory packages/azure-ai poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5
+        run: |
+          uv run --directory packages/azure-ai poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5
+          uv run --directory packages/foundry poe integration-tests -n logical --dist worksteal --timeout=120 --session-timeout=900 --timeout_method thread --retries 2 --retry-delay 5
        working-directory: ./python
      - name: Test Azure AI samples
        timeout-minutes: 10
@@ -78,6 +78,7 @@ jobs:
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
      # GitHub MCP
      GITHUB_PAT: ${{ secrets.GITHUB_TOKEN }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
      # Observability
      ENABLE_INSTRUMENTATION: "true"
    defaults:
@@ -346,7 +347,7 @@ jobs:

  validate-02-agents-amazon:
    name: Validate 02-agents/providers/amazon
-    if: false  # Temporarily disabled - requires AWS credentials
+    if: false # Temporarily disabled - requires AWS credentials
    runs-on: ubuntu-latest
    environment: integration
    env:
@@ -378,7 +379,7 @@ jobs:

  validate-02-agents-ollama:
    name: Validate 02-agents/providers/ollama
-    if: false  # Temporarily disabled - requires local Ollama server
+    if: false # Temporarily disabled - requires local Ollama server
    runs-on: ubuntu-latest
    environment: integration
    env:
@@ -410,7 +411,7 @@ jobs:

  validate-02-agents-foundry-local:
    name: Validate 02-agents/providers/foundry_local
-    if: false  # Temporarily disabled - requires local Foundry setup
+    if: false # Temporarily disabled - requires local Foundry setup
    runs-on: ubuntu-latest
    environment: integration
    defaults:
@@ -440,7 +441,7 @@ jobs:

  validate-02-agents-copilotstudio:
    name: Validate 02-agents/providers/copilotstudio
-    if: false  # Temporarily disabled - requires Copilot Studio setup
+    if: false # Temporarily disabled - requires Copilot Studio setup
    runs-on: ubuntu-latest
    environment: integration
    env:
@@ -556,7 +557,7 @@ jobs:

  validate-04-hosting:
    name: Validate 04-hosting
-    if: false  # Temporarily disabled because of sample complexity
+    if: false # Temporarily disabled because of sample complexity
    runs-on: ubuntu-latest
    environment: integration
    env:
@@ -595,7 +596,7 @@ jobs:

  validate-05-end-to-end:
    name: Validate 05-end-to-end
-    if: false  # Temporarily disabled because of sample complexity
+    if: false # Temporarily disabled because of sample complexity
    runs-on: ubuntu-latest
    environment: integration
    env:
@@ -652,6 +653,7 @@ jobs:
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
    defaults:
      run:
        working-directory: python
@@ -703,6 +705,7 @@ jobs:
      OPENAI_API_KEY: ${{ secrets.OPENAI__APIKEY }}
      OPENAI_CHAT_MODEL_ID: ${{ vars.OPENAI__CHATMODELID }}
      OPENAI_RESPONSES_MODEL_ID: ${{ vars.OPENAI__RESPONSESMODELID }}
+      OPENAI_MODEL: ${{ vars.OPENAI__RESPONSESMODELID }}
      # Copilot Studio
      COPILOTSTUDIOAGENT__ENVIRONMENTID: ${{ secrets.COPILOTSTUDIOAGENT__ENVIRONMENTID }}
      COPILOTSTUDIOAGENT__SCHEMANAME: ${{ secrets.COPILOTSTUDIOAGENT__SCHEMANAME }}
@@ -0,0 +1,72 @@
+---
+status: accepted
+contact: eavanvalkenburg
+date: 2026-03-20
+deciders: eavanvalkenburg, sphenry, chetantoshnival
+consulted: taochenosu, moonbox3, dmytrostruk, giles17, alliscode
+---
+
+# Provider-Leading Client Design & OpenAI Package Extraction
+
+## Context and Problem Statement
+
+The `agent-framework-core` package currently bundles OpenAI and Azure OpenAI client implementations along with their dependencies (`openai`, `azure-identity`, `azure-ai-projects`, `packaging`). This makes core heavier than necessary for users who don't use OpenAI, and it conflates the core abstractions with a specific provider implementation. Additionally, the current class naming (`OpenAIResponsesClient`, `OpenAIChatClient`) is based on the underlying OpenAI API names rather than what users actually want to do, making discoverability harder for newcomers.
+
+## Decision Drivers
+
+- **Lightweight core**: Core should only contain abstractions, middleware infrastructure, and telemetry — no provider-specific code or dependencies.
+- **Discoverability-first**: Import namespaces should guide users to the right client. `from agent_framework.openai import ...` should surface all OpenAI-related clients; `from agent_framework.azure import ...` should surface Foundry, Azure AI, and other Azure-specific classes.
+- **Provider-leading naming**: The primary client name should reflect the provider, not the underlying API. The Responses API is now the recommended default for OpenAI, so its client should be called `OpenAIChatClient` (not `OpenAIResponsesClient`).
+- **Clean separation of concerns**: Azure-specific deprecated wrappers belong in the azure-ai package, not in the OpenAI package.
+
+## Considered Options
+
+- **Keep OpenAI in core**: Simpler but keeps core heavy; doesn't help discoverability.
+- **Extract OpenAI with Azure wrappers in the OpenAI package**: Keeps Azure OpenAI wrappers alongside OpenAI code, but pollutes the OpenAI package with Azure concerns.
+- **Extract OpenAI, place Azure wrappers in azure-ai**: Clean separation; the OpenAI package has zero Azure dependencies; deprecated Azure wrappers live in a single file in azure-ai for easy future deletion.
+
+## Decision Outcome
+
+Chosen option: "Extract OpenAI, place Azure wrappers in azure-ai", because it achieves the lightest core, cleanest OpenAI package, and the most maintainable deprecation path.
+
+Key changes:
+
+1. **New `agent-framework-openai` package** with dependencies on `agent-framework-core`, `openai`, and `packaging` only.
+2. **Class renames**: `OpenAIResponsesClient` → `OpenAIChatClient` (Responses API), `OpenAIChatClient` → `OpenAIChatCompletionClient` (Chat Completions API). Old names remain as deprecated aliases.
+3. **Deprecated classes**: `OpenAIAssistantsClient`, all `AzureOpenAI*Client` classes, `AzureAIClient`, `AzureAIAgentClient`, and `AzureAIProjectAgentProvider` are marked deprecated.
+4. **New `FoundryChatClient`** in azure-ai for Azure AI Foundry Responses API access, built on `RawFoundryChatClient(RawOpenAIChatClient)`.
+5. **All deprecated `AzureOpenAI*` classes** consolidated into a single file (`_deprecated_azure_openai.py`) in the azure-ai package for clean future deletion.
+6. **Core's `agent_framework.openai` and `agent_framework.azure` namespaces** become lazy-loading gateways, preserving backward-compatible import paths while removing hard dependencies.
+7. **Unified `model` parameter** replaces `model_id` (OpenAI), `deployment_name` (Azure OpenAI), and `model_deployment_name` (Azure AI) across all client constructors. The term `model` is intentionally generic: it naturally maps to an OpenAI model name *and* to an Azure OpenAI deployment name, making it straightforward to use `OpenAIChatClient` with either OpenAI or Azure OpenAI backends (via `AsyncAzureOpenAI`). Environment variables are similarly unified (e.g., `OPENAI_MODEL` instead of separate `OPENAI_RESPONSES_MODEL_ID` / `OPENAI_CHAT_MODEL_ID`).
+8. **`FoundryAgent`** replaces the pattern of `Agent(client=AzureAIClient(...))` for connecting to pre-configured agents in Azure AI Foundry (PromptAgents and HostedAgents). The underlying `RawFoundryAgentChatClient` is an implementation detail — most users interact only with `FoundryAgent`. `AzureAIAgentClient` is separately deprecated as it refers to the V1 Agents Service API. See below for design rationale.
+
+### Foundry Agent Design: `FoundryAgentClient` vs `FoundryAgent`
+
+The existing `AzureAIClient` combines two concerns: CRUD lifecycle management (creating/deleting agents on the service) and runtime communication (sending messages via the Responses API). The new design removes CRUD entirely — users connect to agents that already exist in Foundry.
+
+**Two approaches were considered:**
+
+**Option A — `FoundryAgentClient` only (public ChatClient):**
+Users compose `Agent(client=FoundryAgentClient(...), tools=[...])`. This follows the universal `Agent(client=X)` pattern used by every other provider. However, a "client" that wraps a named remote agent (with `agent_name` as a constructor param) is semantically odd — clients typically wrap a model endpoint, not a specific agent.
+
+**Option B — `FoundryAgent` (Agent subclass) + private `_FoundryAgentChatClient` and public `RawFoundryAgentChatClient`:**
+Users write `FoundryAgent(agent_name="my-agent", ...)` for the common case. Internally, `FoundryAgent` creates a `_FoundryAgentChatClient` and passes it to the standard `Agent` base class. For advanced customization, users pass `client_type=RawFoundryAgentChatClient` (or a custom subclass) to control the client middleware layers. The `Agent(client=RawFoundryAgentChatClient(...))` composition pattern still works for users who prefer it.
+
+**Chosen option: Option B**, because:
+- The common case (`FoundryAgent(...)`) is a single object with no boilerplate.
+- `client_type=` gives full control over client middleware without parameter duplication — the agent forwards connection params to the client internally.
+- `RawFoundryAgent(RawAgent)` and `FoundryAgent(Agent)` mirror the established `RawAgent`/`Agent` pattern.
+- Runtime validation (only `FunctionTool` allowed) lives in `RawFoundryAgentChatClient._prepare_options`, ensuring it applies regardless of how the client is used — through `FoundryAgent`, `Agent(client=...)`, or any custom composition.
+
+**Public classes:**
+- `RawFoundryAgentChatClient(RawOpenAIChatClient)` — Responses API client that injects agent reference and validates tools. Extension point for custom client middleware.
+- `RawFoundryAgent(RawAgent)` — Agent without agent-level middleware/telemetry.
+- `FoundryAgent(AgentTelemetryLayer, AgentMiddlewareLayer, RawFoundryAgent)` — Recommended production agent.
+
+**Internal (private):**
+- `_FoundryAgentChatClient` — Full client with function invocation, chat middleware, and telemetry layers. Created automatically by `FoundryAgent`; users customize via `client_type=RawFoundryAgentChatClient` or a custom subclass.
+
+**Deprecated:**
+- `AzureAIClient` — replaced by `FoundryAgent` (which uses `FoundryAgentClient` internally).
+- `AzureAIAgentClient` — refers to V1 Agents Service API, no direct replacement.
+- `AzureAIProjectAgentProvider` — replaced by `FoundryAgent`.
@@ -57,6 +57,7 @@
    <Project Path="samples/02-agents/Agents/Agent_Step16_Declarative/Agent_Step16_Declarative.csproj" />
    <Project Path="samples/02-agents/Agents/Agent_Step17_AdditionalAIContext/Agent_Step17_AdditionalAIContext.csproj" />
    <Project Path="samples/02-agents/Agents/Agent_Step18_CompactionPipeline/Agent_Step18_CompactionPipeline.csproj" />
+    <Project Path="samples/02-agents/Agents/Agent_Step19_InFunctionLoopCheckpointing/Agent_Step19_InFunctionLoopCheckpointing.csproj" />
  </Folder>
  <Folder Name="/Samples/02-agents/DeclarativeAgents/">
    <Project Path="samples/02-agents/DeclarativeAgents/ChatClient/DeclarativeChatClientAgents.csproj" />
@@ -76,6 +77,8 @@
    <Project Path="samples/04-hosting/DurableWorkflows/AzureFunctions/01_SequentialWorkflow/01_SequentialWorkflow.csproj" />
    <Project Path="samples/04-hosting/DurableWorkflows/AzureFunctions/02_ConcurrentWorkflow/02_ConcurrentWorkflow.csproj" />
    <Project Path="samples/04-hosting/DurableWorkflows/AzureFunctions/03_WorkflowHITL/03_WorkflowHITL.csproj" />
+    <Project Path="samples/04-hosting/DurableWorkflows/AzureFunctions/04_WorkflowMcpTool/04_WorkflowMcpTool.csproj" />
+    <Project Path="samples/04-hosting/DurableWorkflows/AzureFunctions/05_WorkflowAndAgents/05_WorkflowAndAgents.csproj" />
  </Folder>
  <Folder Name="/Samples/GettingStarted/">
    <File Path="samples/GettingStarted/README.md" />
@@ -0,0 +1,20 @@
+<Project Sdk="Microsoft.NET.Sdk">
+
+  <PropertyGroup>
+    <OutputType>Exe</OutputType>
+    <TargetFrameworks>net10.0</TargetFrameworks>
+
+    <Nullable>enable</Nullable>
+    <ImplicitUsings>enable</ImplicitUsings>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <PackageReference Include="Azure.AI.OpenAI" />
+    <PackageReference Include="Azure.Identity" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <ProjectReference Include="..\..\..\..\src\Microsoft.Agents.AI.OpenAI\Microsoft.Agents.AI.OpenAI.csproj" />
+  </ItemGroup>
+
+</Project>
@@ -0,0 +1,226 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+// This sample demonstrates how the ChatClientAgent persists chat history after each individual
+// call to the AI service.
+// When an agent uses tools, FunctionInvokingChatClient may loop multiple times
+// (service call → tool execution → service call), and intermediate messages (tool calls and
+// results) are persisted after each service call. This allows you to inspect or recover them
+// even if the process is interrupted mid-loop, but may also result in chat history that is not
+// yet finalized (e.g., tool calls without results) being persisted, which may be undesirable in some cases.
+//
+// To opt into end-of-run persistence instead (atomic run semantics), set
+// PersistChatHistoryAtEndOfRun = true on ChatClientAgentOptions.
+//
+// The sample runs two multi-turn conversations: one using non-streaming (RunAsync) and one
+// using streaming (RunStreamingAsync), to demonstrate correct behavior in both modes.
+
+using System.ComponentModel;
+using Azure.AI.OpenAI;
+using Azure.Identity;
+using Microsoft.Agents.AI;
+using Microsoft.Extensions.AI;
+using OpenAI.Responses;
+
+var endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT") ?? throw new InvalidOperationException("AZURE_OPENAI_ENDPOINT is not set.");
+var deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT_NAME") ?? "gpt-4o-mini";
+var store = Environment.GetEnvironmentVariable("AZURE_OPENAI_RESPONSES_STORE") ?? "false";
+
+// WARNING: DefaultAzureCredential is convenient for development but requires careful consideration in production.
+// In production, consider using a specific credential (e.g., ManagedIdentityCredential) to avoid
+// latency issues, unintended credential probing, and potential security risks from fallback mechanisms.
+AzureOpenAIClient openAIClient = new(new Uri(endpoint), new DefaultAzureCredential());
+
+// Define multiple tools so the model makes several tool calls in a single run.
+[Description("Get the current weather for a city.")]
+static string GetWeather([Description("The city name.")] string city) =>
+    city.ToUpperInvariant() switch
+    {
+        "SEATTLE" => "Seattle: 55°F, cloudy with light rain.",
+        "NEW YORK" => "New York: 72°F, sunny and warm.",
+        "LONDON" => "London: 48°F, overcast with fog.",
+        "DUBLIN" => "Dublin: 43°F, overcast with fog.",
+        _ => $"{city}: weather data not available."
+    };
+
+[Description("Get the current time in a city.")]
+static string GetTime([Description("The city name.")] string city) =>
+    city.ToUpperInvariant() switch
+    {
+        "SEATTLE" => "Seattle: 9:00 AM PST",
+        "NEW YORK" => "New York: 12:00 PM EST",
+        "LONDON" => "London: 5:00 PM GMT",
+        "DUBLIN" => "Dublin: 5:00 PM GMT",
+        _ => $"{city}: time data not available."
+    };
+
+// Create the agent — per-service-call persistence is the default behavior.
+// The in-memory ChatHistoryProvider is used by default when the service does not require service stored chat
+// history, so for those cases, we can inspect the chat history via session.TryGetInMemoryChatHistory().
+IChatClient chatClient = string.Equals(store, "TRUE", StringComparison.OrdinalIgnoreCase) ?
+    openAIClient.GetResponsesClient().AsIChatClient(deploymentName) :
+    openAIClient.GetResponsesClient().AsIChatClientWithStoredOutputDisabled(deploymentName);
+AIAgent agent = chatClient.AsAIAgent(
+    new ChatClientAgentOptions
+    {
+        Name = "WeatherAssistant",
+        ChatOptions = new()
+        {
+            Instructions = "You are a helpful assistant. When asked about multiple cities, call the appropriate tool for each city.",
+            Tools = [AIFunctionFactory.Create(GetWeather), AIFunctionFactory.Create(GetTime)]
+        },
+    });
+
+await RunNonStreamingAsync();
+await RunStreamingAsync();
+
+async Task RunNonStreamingAsync()
+{
+    int lastChatHistorySize = 0;
+    string lastConversationId = string.Empty;
+
+    Console.ForegroundColor = ConsoleColor.Yellow;
+    Console.WriteLine("\n=== Non-Streaming Mode ===");
+    Console.ResetColor();
+
+    AgentSession session = await agent.CreateSessionAsync();
+
+    // First turn — ask about multiple cities so the model calls tools.
+    const string Prompt = "What's the weather and time in Seattle, New York, and London?";
+    PrintUserMessage(Prompt);
+
+    var response = await agent.RunAsync(Prompt, session);
+    PrintAgentResponse(response.Text);
+    PrintChatHistory(session, "After run", ref lastChatHistorySize, ref lastConversationId);
+
+    // Second turn — follow-up to verify chat history is correct.
+    const string FollowUp1 = "And Dublin?";
+    PrintUserMessage(FollowUp1);
+
+    response = await agent.RunAsync(FollowUp1, session);
+    PrintAgentResponse(response.Text);
+    PrintChatHistory(session, "After second run", ref lastChatHistorySize, ref lastConversationId);
+
+    // Third turn — follow-up to verify chat history is correct.
+    const string FollowUp2 = "Which city is the warmest?";
+    PrintUserMessage(FollowUp2);
+
+    response = await agent.RunAsync(FollowUp2, session);
+    PrintAgentResponse(response.Text);
+    PrintChatHistory(session, "After third run", ref lastChatHistorySize, ref lastConversationId);
+}
+
+async Task RunStreamingAsync()
+{
+    int lastChatHistorySize = 0;
+    string lastConversationId = string.Empty;
+
+    Console.ForegroundColor = ConsoleColor.Yellow;
+    Console.WriteLine("\n=== Streaming Mode ===");
+    Console.ResetColor();
+
+    AgentSession session = await agent.CreateSessionAsync();
+
+    // First turn — ask about multiple cities so the model calls tools.
+    const string Prompt = "What's the weather and time in Seattle, New York, and London?";
+    PrintUserMessage(Prompt);
+
+    Console.ForegroundColor = ConsoleColor.Cyan;
+    Console.Write("\n[Agent] ");
+    Console.ResetColor();
+
+    await foreach (var update in agent.RunStreamingAsync(Prompt, session))
+    {
+        Console.Write(update);
+
+        // During streaming we should be able to see updates to the chat history
+        // before the full run completes, as each service call is made and persisted.
+        PrintChatHistory(session, "During run", ref lastChatHistorySize, ref lastConversationId);
+    }
+
+    Console.WriteLine();
+    PrintChatHistory(session, "After run", ref lastChatHistorySize, ref lastConversationId);
+
+    // Second turn — follow-up to verify chat history is correct.
+    const string FollowUp1 = "And Dublin?";
+    PrintUserMessage(FollowUp1);
+
+    Console.ForegroundColor = ConsoleColor.Cyan;
+    Console.Write("\n[Agent] ");
+    Console.ResetColor();
+
+    await foreach (var update in agent.RunStreamingAsync(FollowUp1, session))
+    {
+        Console.Write(update);
+
+        // During streaming we should be able to see updates to the chat history
+        // before the full run completes, as each service call is made and persisted.
+        PrintChatHistory(session, "During second run", ref lastChatHistorySize, ref lastConversationId);
+    }
+
+    Console.WriteLine();
+    PrintChatHistory(session, "After second run", ref lastChatHistorySize, ref lastConversationId);
+
+    // Third turn — follow-up to verify chat history is correct.
+    const string FollowUp2 = "Which city is the warmest?";
+    PrintUserMessage(FollowUp2);
+
+    Console.ForegroundColor = ConsoleColor.Cyan;
+    Console.Write("\n[Agent] ");
+    Console.ResetColor();
+
+    await foreach (var update in agent.RunStreamingAsync(FollowUp2, session))
+    {
+        Console.Write(update);
+
+        // During streaming we should be able to see updates to the chat history
+        // before the full run completes, as each service call is made and persisted.
+        PrintChatHistory(session, "During third run", ref lastChatHistorySize, ref lastConversationId);
+    }
+
+    Console.WriteLine();
+    PrintChatHistory(session, "After third run", ref lastChatHistorySize, ref lastConversationId);
+}
+
+void PrintUserMessage(string message)
+{
+    Console.ForegroundColor = ConsoleColor.Cyan;
+    Console.Write("\n[User] ");
+    Console.ResetColor();
+    Console.WriteLine(message);
+}
+
+void PrintAgentResponse(string? text)
+{
+    Console.ForegroundColor = ConsoleColor.Cyan;
+    Console.Write("\n[Agent] ");
+    Console.ResetColor();
+    Console.WriteLine(text);
+}
+
+// Helper to print the current chat history from the session.
+void PrintChatHistory(AgentSession session, string label, ref int lastChatHistorySize, ref string lastConversationId)
+{
+    if (session.TryGetInMemoryChatHistory(out var history) && history.Count != lastChatHistorySize)
+    {
+        Console.ForegroundColor = ConsoleColor.DarkGray;
+        Console.WriteLine($"\n  [{label} — Chat history: {history.Count} message(s)]");
+        foreach (var msg in history)
+        {
+            var preview = msg.Text?.Length > 80 ? msg.Text[..80] + "…" : msg.Text;
+            var contentTypes = string.Join(", ", msg.Contents.Select(c => c.GetType().Name));
+            Console.WriteLine($"    {msg.Role,-12} | {(string.IsNullOrWhiteSpace(preview) ? $"[{contentTypes}]" : preview)}");
+        }
+
+        Console.ResetColor();
+
+        lastChatHistorySize = history.Count;
+    }
+
+    if (session is ChatClientAgentSession ccaSession && ccaSession.ConversationId is not null && ccaSession.ConversationId != lastConversationId)
+    {
+        Console.ForegroundColor = ConsoleColor.DarkGray;
+        Console.WriteLine($"  [{label} — Conversation ID: {ccaSession.ConversationId}]");
+        Console.ResetColor();
+        lastConversationId = ccaSession.ConversationId;
+    }
+}
@@ -0,0 +1,63 @@
+# In-Function-Loop Checkpointing
+
+This sample demonstrates how `ChatClientAgent` persists chat history after each individual call to the AI service by default. This per-service-call persistence ensures intermediate progress is saved during the function invocation loop.
+
+## What This Sample Shows
+
+When an agent uses tools, the `FunctionInvokingChatClient` loops multiple times (service call → tool execution → service call → …). By default, chat history is persisted after each service call via the `ChatHistoryPersistingChatClient` decorator:
+
+- A `ChatHistoryPersistingChatClient` decorator is automatically inserted into the chat client pipeline
+- After each service call, the decorator notifies the `ChatHistoryProvider` (and any `AIContextProvider` instances) with the new messages
+- Only **new** messages are sent to providers on each notification — messages that were already persisted in an earlier call within the same run are deduplicated automatically
+
+To opt into end-of-run persistence instead (atomic run semantics), set `PersistChatHistoryAtEndOfRun = true` on `ChatClientAgentOptions`. In that mode, the decorator marks messages with metadata rather than persisting them immediately, and `ChatClientAgent` persists only the marked messages at the end of the run.
+
+Per-service-call persistence is useful for:
+- **Crash recovery** — if the process is interrupted mid-loop, the intermediate tool calls and results are already persisted
+- **Observability** — you can inspect the chat history while the agent is still running (e.g., during streaming)
+- **Long-running tool loops** — agents with many sequential tool calls benefit from incremental persistence
+
+## How It Works
+
+The sample asks the agent about the weather and time in three cities. The model calls the `GetWeather` and `GetTime` tools for each city, resulting in multiple service calls within a single `RunStreamingAsync` invocation. After the run completes, the sample prints the full chat history to show all the intermediate messages that were persisted along the way.
+
+### Pipeline Architecture
+
+```
+ChatClientAgent
+  └─ FunctionInvokingChatClient    (handles tool call loop)
+       └─ ChatHistoryPersistingChatClient  (persists after each service call)
+            └─ Leaf IChatClient            (Azure OpenAI)
+```
+
+## Prerequisites
+
+- .NET 10 SDK or later
+- Azure OpenAI service endpoint and model deployment
+- Azure CLI installed and authenticated
+
+**Note**: This sample uses `DefaultAzureCredential`. Sign in with `az login` before running. For production, prefer a specific credential such as `ManagedIdentityCredential`. For more information, see the [Azure CLI authentication documentation](https://learn.microsoft.com/cli/azure/authenticate-azure-cli-interactively).
+
+## Environment Variables
+
+```powershell
+$env:AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com/"  # Required
+$env:AZURE_OPENAI_DEPLOYMENT_NAME="gpt-4o-mini"                       # Optional, defaults to gpt-4o-mini
+```
+
+## Running the Sample
+
+```powershell
+cd dotnet/samples/02-agents/Agents/Agent_Step19_InFunctionLoopCheckpointing
+dotnet run
+```
+
+## Expected Behavior
+
+The sample runs two conversation turns:
+
+1. **First turn** — asks about weather and time in three cities. The model calls `GetWeather` and `GetTime` tools (potentially in parallel or sequentially), then provides a summary. The chat history dump after the run shows all the intermediate tool call and result messages.
+
+2. **Second turn** — asks a follow-up question ("Which city is the warmest?") that uses the persisted conversation context. The chat history dump shows the full accumulated conversation.
+
+The chat history printout uses `session.TryGetInMemoryChatHistory()` to inspect the in-memory storage.
@@ -45,6 +45,7 @@ Before you begin, ensure you have the following prerequisites:
 |[Declarative agent](./Agent_Step16_Declarative/)|This sample demonstrates how to declaratively define an agent.|
 |[Providing additional AI Context to an agent using multiple AIContextProviders](./Agent_Step17_AdditionalAIContext/)|This sample demonstrates how to inject additional AI context into a ChatClientAgent using multiple custom AIContextProvider components that are attached to the agent.|
 |[Using compaction pipeline with an agent](./Agent_Step18_CompactionPipeline/)|This sample demonstrates how to use a compaction pipeline to efficiently limit the size of the conversation history for an agent.|
+|[In-function-loop checkpointing](./Agent_Step19_InFunctionLoopCheckpointing/)|This sample demonstrates how to persist chat history after each service call during a tool-calling loop, enabling crash recovery and mid-run observability.|

 ## Running the samples from the console

@@ -9,7 +9,7 @@ using Microsoft.Extensions.AI;
 namespace WorkflowAsAnAgentSample;

 /// <summary>
-/// This sample introduces the concepts workflows as agents, where a workflow can be
+/// This sample introduces the concept of workflows as agents, where a workflow can be
 /// treated as an <see cref="AIAgent"/>. This allows you to interact with a workflow
 /// as if it were a single agent.
 ///
@@ -18,6 +18,14 @@ namespace WorkflowAsAnAgentSample;
 ///
 /// You will interact with the workflow in an interactive loop, sending messages and receiving
 /// streaming responses from the workflow as if it were an agent who responds in both languages.
+///
+/// This sample also demonstrates <see cref="IResettableExecutor"/>, which is required
+/// for stateful executors that are shared across multiple workflow runs. Each iteration
+/// of the interactive loop triggers a new workflow run against the same workflow instance.
+/// Between runs, the framework automatically calls <see cref="IResettableExecutor.ResetAsync"/>
+/// on shared executors so that accumulated state (e.g., collected messages) is cleared
+/// before the next run begins. See <c>WorkflowFactory.ConcurrentAggregationExecutor</c>
+/// for the implementation.
 /// </summary>
 /// <remarks>
 /// Pre-requisites:
@@ -39,7 +47,10 @@ public static class Program
        var agent = workflow.AsAIAgent("workflow-agent", "Workflow Agent");
        var session = await agent.CreateSessionAsync();

-        // Start an interactive loop to interact with the workflow as if it were an agent
+        // Start an interactive loop to interact with the workflow as if it were an agent.
+        // Each iteration runs the workflow again on the same workflow instance. Between runs,
+        // the framework calls IResettableExecutor.ResetAsync() on shared stateful executors
+        // (like ConcurrentAggregationExecutor) to clear accumulated state from the previous run.
        while (true)
        {
            Console.WriteLine();
@@ -10,6 +10,14 @@ internal static class WorkflowFactory
 {
    /// <summary>
    /// Creates a workflow that uses two language agents to process input concurrently.
+    ///
+    /// In this workflow, the <c>Start</c> <see cref="ChatForwardingExecutor"/> and the
+    /// <see cref="ConcurrentAggregationExecutor"/> are provided as shared instances, meaning
+    /// the same executor objects are reused across multiple workflow runs. The language agents
+    /// (French and English) are created via a factory and instantiated per workflow run.
+    /// Stateful shared executors must implement <see cref="IResettableExecutor"/> so the
+    /// framework can clear their state between runs. Framework-provided executors like
+    /// <see cref="ChatForwardingExecutor"/> already implement this interface.
    /// </summary>
    /// <param name="chatClient">The chat client to use for the agents</param>
    /// <returns>A workflow that processes input using two language agents</returns>
@@ -40,6 +48,16 @@ internal static class WorkflowFactory

    /// <summary>
    /// Executor that aggregates the results from the concurrent agents.
+    ///
+    /// This executor is stateful — it accumulates messages in <see cref="_messages"/>
+    /// as they arrive from each agent. Because it is provided as a shared instance
+    /// (not via a factory), the same object is reused across workflow runs. Implementing
+    /// <see cref="IResettableExecutor"/> allows the framework to call <see cref="ResetAsync"/>
+    /// between runs, clearing accumulated state so each run starts fresh.
+    ///
+    /// Without <see cref="IResettableExecutor"/>, attempting to reuse a workflow containing
+    /// shared executor instances that do not implement this interface would throw an
+    /// <see cref="InvalidOperationException"/>.
    /// </summary>
    [YieldsOutput(typeof(string))]
    private sealed class ConcurrentAggregationExecutor() :
@@ -65,7 +83,11 @@ internal static class WorkflowFactory
            }
        }

-        /// <inheritdoc/>
+        /// <summary>
+        /// Resets the executor state between workflow runs by clearing accumulated messages.
+        /// The framework calls this automatically when a workflow run completes, before the
+        /// workflow can be used for another run.
+        /// </summary>
        public ValueTask ResetAsync()
        {
            this._messages.Clear();
@@ -0,0 +1,35 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <TargetFrameworks>net10.0</TargetFrameworks>
+    <AzureFunctionsVersion>v4</AzureFunctionsVersion>
+    <OutputType>Exe</OutputType>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+    <!-- The Functions build tools don't like namespaces that start with a number -->
+    <AssemblyName>WorkflowMcpTool</AssemblyName>
+    <RootNamespace>WorkflowMcpTool</RootNamespace>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <FrameworkReference Include="Microsoft.AspNetCore.App" />
+  </ItemGroup>
+
+  <!-- Azure Functions packages -->
+  <ItemGroup>
+    <PackageReference Include="Microsoft.Azure.Functions.Worker" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.DurableTask" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.DurableTask.AzureManaged" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.Http.AspNetCore" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Sdk" />
+  </ItemGroup>
+
+  <!-- Local projects that should be switched to package references when using the sample outside of this MAF repo -->
+  <!--
+  <ItemGroup>
+    <PackageReference Include="Microsoft.Agents.AI.Hosting.AzureFunctions" />
+  </ItemGroup>
+  -->
+  <ItemGroup>
+    <ProjectReference Include="..\..\..\..\..\src\Microsoft.Agents.AI.Hosting.AzureFunctions\Microsoft.Agents.AI.Hosting.AzureFunctions.csproj" />
+  </ItemGroup>
+</Project>
@@ -0,0 +1,59 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using Microsoft.Agents.AI.Workflows;
+
+namespace WorkflowMcpTool;
+
+internal sealed class TranslateText() : Executor<string, TranslationResult>("TranslateText")
+{
+    public override ValueTask<TranslationResult> HandleAsync(
+        string message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine($"[Activity] TranslateText: '{message}'");
+        return ValueTask.FromResult(new TranslationResult(message, message.ToUpperInvariant()));
+    }
+}
+
+internal sealed class FormatOutput() : Executor<TranslationResult, string>("FormatOutput")
+{
+    public override ValueTask<string> HandleAsync(
+        TranslationResult message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine("[Activity] FormatOutput: Formatting result");
+        return ValueTask.FromResult($"Original: {message.Original} => Translated: {message.Translated}");
+    }
+}
+
+internal sealed class LookupOrder() : Executor<string, OrderInfo>("LookupOrder")
+{
+    public override ValueTask<OrderInfo> HandleAsync(
+        string message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine($"[Activity] LookupOrder: '{message}'");
+        return ValueTask.FromResult(new OrderInfo(message, "Alice Johnson", "Wireless Headphones", Quantity: 2, UnitPrice: 49.99m));
+    }
+}
+
+internal sealed class EnrichOrder() : Executor<OrderInfo, OrderSummary>("EnrichOrder")
+{
+    public override ValueTask<OrderSummary> HandleAsync(
+        OrderInfo message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine($"[Activity] EnrichOrder: '{message.OrderId}'");
+        return ValueTask.FromResult(new OrderSummary(message, TotalPrice: message.Quantity * message.UnitPrice, Status: "Confirmed"));
+    }
+}
+
+internal sealed record TranslationResult(string Original, string Translated);
+
+internal sealed record OrderInfo(string OrderId, string CustomerName, string Product, int Quantity, decimal UnitPrice);
+
+internal sealed record OrderSummary(OrderInfo Order, decimal TotalPrice, string Status);
@@ -0,0 +1,44 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+// This sample demonstrates how to expose a durable workflow as an MCP (Model Context Protocol) tool.
+// When using AddWorkflow with exposeMcpToolTrigger: true, the Functions host will automatically
+// generate a remote MCP endpoint for the app at /runtime/webhooks/mcp with a workflow-specific
+// tool name. MCP-compatible clients can then invoke the workflow as a tool.
+
+using Microsoft.Agents.AI.Hosting.AzureFunctions;
+using Microsoft.Agents.AI.Workflows;
+using Microsoft.Azure.Functions.Worker.Builder;
+using Microsoft.Extensions.Hosting;
+using WorkflowMcpTool;
+
+// Define executors
+TranslateText translateText = new();
+FormatOutput formatOutput = new();
+LookupOrder lookupOrder = new();
+EnrichOrder enrichOrder = new();
+
+// Build a simple workflow: TranslateText -> FormatOutput
+Workflow translateWorkflow = new WorkflowBuilder(translateText)
+    .WithName("Translate")
+    .WithDescription("Translate text to uppercase and format the result")
+    .AddEdge(translateText, formatOutput)
+    .Build();
+
+// Build a workflow that returns a POCO: LookupOrder -> EnrichOrder
+Workflow orderLookupWorkflow = new WorkflowBuilder(lookupOrder)
+    .WithName("OrderLookup")
+    .WithDescription("Look up an order by ID and return enriched order details")
+    .AddEdge(lookupOrder, enrichOrder)
+    .Build();
+
+using IHost app = FunctionsApplication
+    .CreateBuilder(args)
+    .ConfigureFunctionsWebApplication()
+    .ConfigureDurableWorkflows(workflows =>
+    {
+        // Expose both workflows as MCP tool triggers.
+        workflows.AddWorkflow(translateWorkflow, exposeStatusEndpoint: false, exposeMcpToolTrigger: true);
+        workflows.AddWorkflow(orderLookupWorkflow, exposeStatusEndpoint: false, exposeMcpToolTrigger: true);
+    })
+    .Build();
+app.Run();
@@ -0,0 +1,81 @@
+# Workflow as MCP Tool Sample
+
+This sample demonstrates how to expose durable workflows as [MCP (Model Context Protocol)](https://modelcontextprotocol.io/) tools, enabling MCP-compatible clients to invoke workflows directly.
+
+## Key Concepts Demonstrated
+
+- **Workflow as MCP Tool**: Expose workflows as callable MCP tools using `exposeMcpToolTrigger: true`
+- **MCP Server Hosting**: The Azure Functions host automatically generates a remote MCP endpoint at `/runtime/webhooks/mcp`
+- **String and POCO Results**: Shows workflows returning both plain strings and structured JSON objects
+
+## Sample Architecture
+
+The sample creates two workflows exposed as MCP tools:
+
+### Translate Workflow (returns a string)
+
+| Executor | Input | Output | Description |
+|----------|-------|--------|-------------|
+| **TranslateText** | `string` | `TranslationResult` | Converts input text to uppercase |
+| **FormatOutput** | `TranslationResult` | `string` | Formats the result into a readable string |
+
+### OrderLookup Workflow (returns a POCO)
+
+| Executor | Input | Output | Description |
+|----------|-------|--------|-------------|
+| **LookupOrder** | `string` | `OrderInfo` | Looks up an order by ID |
+| **EnrichOrder** | `OrderInfo` | `OrderSummary` | Adds computed fields (total price, status) |
+
+## Environment Setup
+
+See the [README.md](../../README.md) file in the parent directory for complete setup instructions, including:
+
+- Prerequisites installation
+- Durable Task Scheduler setup
+- Storage emulator configuration
+
+For this sample, you'll also need [Node.js](https://nodejs.org/en/download) to use the [MCP Inspector](https://modelcontextprotocol.io/docs/tools/inspector).
+
+## Running the Sample
+
+1. **Start the Function App**:
+
+   ```bash
+   cd dotnet/samples/04-hosting/DurableWorkflows/AzureFunctions/04_WorkflowMcpTool
+   func start
+   ```
+
+2. **Note the MCP Server Endpoint**: When the app starts, you'll see the MCP server endpoint in the terminal output:
+
+   ```text
+   MCP server endpoint:  http://localhost:7071/runtime/webhooks/mcp
+   ```
+
+## Invoking Workflows via MCP Inspector
+
+1. Install and run the [MCP Inspector](https://modelcontextprotocol.io/docs/tools/inspector):
+
+   ```bash
+   npx @modelcontextprotocol/inspector
+   ```
+
+2. Connect to the MCP server endpoint:
+   - For **Transport Type**, select **"Streamable HTTP"**
+   - For **URL**, enter `http://localhost:7071/runtime/webhooks/mcp`
+   - Click the **Connect** button
+
+3. Click the **List Tools** button. You should see two tools: `Translate` and `OrderLookup`.
+
+4. Test the **Translate** tool (returns a plain string):
+   - Select the `Translate` tool
+   - Set `hello world` as the `input` parameter
+   - Click **Run Tool**
+   - Expected result: `Original: hello world => Translated: HELLO WORLD`
+
+5. Test the **OrderLookup** tool (returns a JSON object):
+   - Select the `OrderLookup` tool
+   - Set `ORD-2025-42` as the `input` parameter
+   - Click **Run Tool**
+   - Expected result: A JSON object containing order details such as `OrderId`, `CustomerName`, `Product`, `TotalPrice`, and `Status`
+
+You'll see the workflow executor activities logged in the terminal where you ran `func start`.
@@ -0,0 +1,20 @@
+{
+  "version": "2.0",
+  "logging": {
+    "logLevel": {
+      "Microsoft.Agents.AI.DurableTask": "Information",
+      "Microsoft.Agents.AI.Hosting.AzureFunctions": "Information",
+      "DurableTask": "Information",
+      "Microsoft.DurableTask": "Information"
+    }
+  },
+  "extensions": {
+    "durableTask": {
+      "hubName": "default",
+      "storageProvider": {
+        "type": "AzureManaged",
+        "connectionStringName": "DURABLE_TASK_SCHEDULER_CONNECTION_STRING"
+      }
+    }
+  }
+}
@@ -0,0 +1,8 @@
+{
+  "IsEncrypted": false,
+  "Values": {
+    "FUNCTIONS_WORKER_RUNTIME": "dotnet-isolated",
+    "AzureWebJobsStorage": "UseDevelopmentStorage=true",
+    "DURABLE_TASK_SCHEDULER_CONNECTION_STRING": "Endpoint=http://localhost:8080;TaskHub=default;Authentication=None"
+  }
+}
@@ -0,0 +1,42 @@
+<Project Sdk="Microsoft.NET.Sdk">
+  <PropertyGroup>
+    <TargetFrameworks>net10.0</TargetFrameworks>
+    <AzureFunctionsVersion>v4</AzureFunctionsVersion>
+    <OutputType>Exe</OutputType>
+    <ImplicitUsings>enable</ImplicitUsings>
+    <Nullable>enable</Nullable>
+    <!-- The Functions build tools don't like namespaces that start with a number -->
+    <AssemblyName>WorkflowAndAgents</AssemblyName>
+    <RootNamespace>WorkflowAndAgents</RootNamespace>
+  </PropertyGroup>
+
+  <ItemGroup>
+    <FrameworkReference Include="Microsoft.AspNetCore.App" />
+  </ItemGroup>
+
+  <!-- Azure Functions packages -->
+  <ItemGroup>
+    <PackageReference Include="Microsoft.Azure.Functions.Worker" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.DurableTask" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.DurableTask.AzureManaged" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Extensions.Http.AspNetCore" />
+    <PackageReference Include="Microsoft.Azure.Functions.Worker.Sdk" />
+  </ItemGroup>
+
+  <ItemGroup>
+    <PackageReference Include="Azure.AI.OpenAI" />
+    <PackageReference Include="Azure.Identity" />
+  </ItemGroup>
+
+  <!-- Local projects that should be switched to package references when using the sample outside of this MAF repo -->
+  <!--
+  <ItemGroup>
+    <PackageReference Include="Microsoft.Agents.AI.Hosting.AzureFunctions" />
+    <PackageReference Include="Microsoft.Agents.AI.OpenAI" />
+  </ItemGroup>
+  -->
+  <ItemGroup>
+    <ProjectReference Include="..\..\..\..\..\src\Microsoft.Agents.AI.Hosting.AzureFunctions\Microsoft.Agents.AI.Hosting.AzureFunctions.csproj" />
+    <ProjectReference Include="..\..\..\..\..\src\Microsoft.Agents.AI.OpenAI\Microsoft.Agents.AI.OpenAI.csproj" />
+  </ItemGroup>
+</Project>
@@ -0,0 +1,31 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using Microsoft.Agents.AI.Workflows;
+
+namespace WorkflowAndAgents;
+
+internal sealed class TranslateText() : Executor<string, TranslationResult>("TranslateText")
+{
+    public override ValueTask<TranslationResult> HandleAsync(
+        string message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine($"[Activity] TranslateText: '{message}'");
+        return ValueTask.FromResult(new TranslationResult(message, message.ToUpperInvariant()));
+    }
+}
+
+internal sealed class FormatOutput() : Executor<TranslationResult, string>("FormatOutput")
+{
+    public override ValueTask<string> HandleAsync(
+        TranslationResult message,
+        IWorkflowContext context,
+        CancellationToken cancellationToken = default)
+    {
+        Console.WriteLine("[Activity] FormatOutput: Formatting result");
+        return ValueTask.FromResult($"Original: {message.Original} => Translated: {message.Translated}");
+    }
+}
+
+internal sealed record TranslationResult(string Original, string Translated);
@@ -0,0 +1,64 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+// This sample demonstrates using ConfigureDurableOptions to register BOTH agents AND workflows
+// in a single Azure Functions app. It uses a workflow to translate text and a standalone AI agent
+// accessible via HTTP and MCP tool triggers.
+
+#pragma warning disable IDE0002 // Simplify Member Access
+
+using Azure;
+using Azure.AI.OpenAI;
+using Azure.Identity;
+using Microsoft.Agents.AI;
+using Microsoft.Agents.AI.Hosting.AzureFunctions;
+using Microsoft.Agents.AI.Workflows;
+using Microsoft.Azure.Functions.Worker.Builder;
+using Microsoft.Extensions.Hosting;
+using OpenAI.Chat;
+using WorkflowAndAgents;
+
+// Get the Azure OpenAI endpoint and deployment name from environment variables.
+string endpoint = Environment.GetEnvironmentVariable("AZURE_OPENAI_ENDPOINT")
+    ?? throw new InvalidOperationException("AZURE_OPENAI_ENDPOINT is not set.");
+string deploymentName = Environment.GetEnvironmentVariable("AZURE_OPENAI_DEPLOYMENT_NAME")
+    ?? throw new InvalidOperationException("AZURE_OPENAI_DEPLOYMENT_NAME is not set.");
+
+// Use Azure Key Credential if provided, otherwise use Azure CLI Credential.
+string? azureOpenAiKey = Environment.GetEnvironmentVariable("AZURE_OPENAI_API_KEY");
+AzureOpenAIClient client = !string.IsNullOrEmpty(azureOpenAiKey)
+    ? new AzureOpenAIClient(new Uri(endpoint), new AzureKeyCredential(azureOpenAiKey))
+    : new AzureOpenAIClient(new Uri(endpoint), new DefaultAzureCredential());
+
+ChatClient chatClient = client.GetChatClient(deploymentName);
+
+// Define a standalone AI agent
+AIAgent assistant = chatClient.AsAIAgent(
+    "You are a helpful assistant. Answer questions clearly and concisely.",
+    "Assistant",
+    description: "A general-purpose helpful assistant.");
+
+// Define workflow executors
+TranslateText translateText = new();
+FormatOutput formatOutput = new();
+
+// Build a workflow: TranslateText -> FormatOutput
+Workflow translateWorkflow = new WorkflowBuilder(translateText)
+    .WithName("Translate")
+    .WithDescription("Translate text to uppercase and format the result")
+    .AddEdge(translateText, formatOutput)
+    .Build();
+
+// Use ConfigureDurableOptions to register both agents and workflows together
+using IHost app = FunctionsApplication
+    .CreateBuilder(args)
+    .ConfigureFunctionsWebApplication()
+    .ConfigureDurableOptions(options =>
+    {
+        // Register the standalone agent with HTTP and MCP tool triggers
+        options.Agents.AddAIAgent(assistant, enableHttpTrigger: true, enableMcpToolTrigger: true);
+
+        // Register the workflow with an HTTP endpoint and MCP tool trigger
+        options.Workflows.AddWorkflow(translateWorkflow, exposeStatusEndpoint: false, exposeMcpToolTrigger: true);
+    })
+    .Build();
+app.Run();
@@ -0,0 +1,76 @@
+# Workflow and Agents Sample
+
+This sample demonstrates how to use `ConfigureDurableOptions` to register **both** AI agents **and** workflows in a single Azure Functions app. This is the recommended approach when your application needs both standalone agents and orchestrated workflows.
+
+## Key Concepts Demonstrated
+
+- **Unified Configuration**: Use `ConfigureDurableOptions` to register agents and workflows together
+- **Standalone Agent**: An AI agent accessible via HTTP and MCP tool triggers
+- **Workflow**: A simple text translation workflow also exposed as an MCP tool
+- **Mixed Triggers**: Both agents and workflows coexist in the same Functions host
+
+## Sample Architecture
+
+### Standalone Agent
+
+| Agent | Description |
+|-------|-------------|
+| **Assistant** | A general-purpose AI assistant accessible via HTTP (`/agents/Assistant/run`) and as an MCP tool |
+
+### Translate Workflow
+
+| Executor | Input | Output | Description |
+|----------|-------|--------|-------------|
+| **TranslateText** | `string` | `TranslationResult` | Converts input text to uppercase |
+| **FormatOutput** | `TranslationResult` | `string` | Formats the result into a readable string |
+
+## Environment Setup
+
+See the [README.md](../../README.md) file in the parent directory for complete setup instructions, including:
+
+- Prerequisites installation
+- Durable Task Scheduler setup
+- Storage emulator configuration
+
+This sample also requires Azure OpenAI credentials. Set the following in `local.settings.json`:
+
+- `AZURE_OPENAI_ENDPOINT`: Your Azure OpenAI endpoint URL
+- `AZURE_OPENAI_DEPLOYMENT_NAME`: Your chat model deployment name
+- `AZURE_OPENAI_API_KEY` (optional): If not set, Azure CLI credential is used
+
+## Running the Sample
+
+1. **Start the Function App**:
+
+   ```bash
+   cd dotnet/samples/04-hosting/DurableWorkflows/AzureFunctions/05_WorkflowAndAgents
+   func start
+   ```
+
+2. **Expected Functions**: When the app starts, you should see functions for both the agent and the workflow:
+
+   - `dafx-Assistant` (entity trigger for the agent)
+   - `http-Assistant` (HTTP trigger for the agent)
+   - `mcptool-Assistant` (MCP tool trigger for the agent)
+   - `wf-Translate` (orchestration trigger for the workflow)
+   - `mcptool-wf-Translate` (MCP tool trigger for the workflow)
+
+## Invoking the Agent via HTTP
+
+```bash
+curl -X POST http://localhost:7071/agents/Assistant/run \
+  -H "Content-Type: application/json" \
+  -d '{"query": "What is the capital of France?"}'
+```
+
+## Invoking via MCP Inspector
+
+1. Install and run the [MCP Inspector](https://modelcontextprotocol.io/docs/tools/inspector):
+
+   ```bash
+   npx @modelcontextprotocol/inspector
+   ```
+
+2. Connect to `http://localhost:7071/runtime/webhooks/mcp` using **Streamable HTTP** transport.
+
+3. Click **List Tools** to see both the `Assistant` agent tool and the `Translate` workflow tool.
@@ -0,0 +1,20 @@
+{
+  "version": "2.0",
+  "logging": {
+    "logLevel": {
+      "Microsoft.Agents.AI.DurableTask": "Information",
+      "Microsoft.Agents.AI.Hosting.AzureFunctions": "Information",
+      "DurableTask": "Information",
+      "Microsoft.DurableTask": "Information"
+    }
+  },
+  "extensions": {
+    "durableTask": {
+      "hubName": "default",
+      "storageProvider": {
+        "type": "AzureManaged",
+        "connectionStringName": "DURABLE_TASK_SCHEDULER_CONNECTION_STRING"
+      }
+    }
+  }
+}
@@ -0,0 +1,10 @@
+{
+  "IsEncrypted": false,
+  "Values": {
+    "FUNCTIONS_WORKER_RUNTIME": "dotnet-isolated",
+    "AzureWebJobsStorage": "UseDevelopmentStorage=true",
+    "DURABLE_TASK_SCHEDULER_CONNECTION_STRING": "Endpoint=http://localhost:8080;TaskHub=default;Authentication=None",
+    "AZURE_OPENAI_ENDPOINT": "<AZURE_OPENAI_ENDPOINT>",
+    "AZURE_OPENAI_DEPLOYMENT_NAME": "<AZURE_OPENAI_DEPLOYMENT_NAME>"
+  }
+}
@@ -48,3 +48,4 @@ $env:DURABLE_TASK_SCHEDULER_CONNECTION_STRING = "AccountEndpoint=http://localhos
 | [01_SequentialWorkflow](AzureFunctions/01_SequentialWorkflow/) | Sequential workflow hosted in Azure Functions |
 | [02_ConcurrentWorkflow](AzureFunctions/02_ConcurrentWorkflow/) | Concurrent workflow hosted in Azure Functions |
 | [03_WorkflowHITL](AzureFunctions/03_WorkflowHITL/) | Human-in-the-loop workflow hosted in Azure Functions |
+| [04_WorkflowMcpTool](AzureFunctions/04_WorkflowMcpTool/) | Workflow exposed as an MCP tool |
@@ -167,6 +167,20 @@ internal sealed class BuiltInFunctionExecutor : IFunctionExecutor
            return;
        }

+        if (context.FunctionDefinition.EntryPoint == BuiltInFunctions.RunWorkflowMcpToolFunctionEntryPoint)
+        {
+            if (mcpToolInvocationContext is null)
+            {
+                throw new InvalidOperationException($"MCP tool invocation context binding is missing for the invocation {context.InvocationId}.");
+            }
+
+            context.GetInvocationResult().Value = await BuiltInFunctions.RunWorkflowMcpToolAsync(
+                mcpToolInvocationContext,
+                durableTaskClient,
+                context);
+            return;
+        }
+
        throw new InvalidOperationException($"Unsupported function entry point '{context.FunctionDefinition.EntryPoint}' for invocation {context.InvocationId}.");
    }

@@ -29,6 +29,7 @@ internal static class BuiltInFunctions
    internal static readonly string InvokeWorkflowActivityFunctionEntryPoint = $"{typeof(BuiltInFunctions).FullName!}.{nameof(InvokeWorkflowActivityAsync)}";
    internal static readonly string GetWorkflowStatusHttpFunctionEntryPoint = $"{typeof(BuiltInFunctions).FullName!}.{nameof(GetWorkflowStatusAsync)}";
    internal static readonly string RespondToWorkflowHttpFunctionEntryPoint = $"{typeof(BuiltInFunctions).FullName!}.{nameof(RespondToWorkflowAsync)}";
+    internal static readonly string RunWorkflowMcpToolFunctionEntryPoint = $"{typeof(BuiltInFunctions).FullName!}.{nameof(RunWorkflowMcpToolAsync)}";

 #pragma warning disable IL3000 // Avoid accessing Assembly file path when publishing as a single file - Azure Functions does not use single-file publishing
    internal static readonly string ScriptFile = Path.GetFileName(typeof(BuiltInFunctions).Assembly.Location);
@@ -378,6 +379,55 @@ internal static class BuiltInFunctions
        return agentResponse.Text;
    }

+    /// <summary>
+    /// Runs a workflow via MCP tool trigger.
+    /// Extracts the <c>input</c> argument, schedules a new orchestration, waits for completion, and returns the output.
+    /// </summary>
+    public static async Task<string?> RunWorkflowMcpToolAsync(
+        [McpToolTrigger("BuiltInWorkflowMcpTool")] ToolInvocationContext context,
+        [DurableClient] DurableTaskClient client,
+        FunctionContext functionContext)
+    {
+        if (context.Arguments is null)
+        {
+            throw new ArgumentException("MCP Tool invocation is missing required arguments.");
+        }
+
+        if (!context.Arguments.TryGetValue("input", out object? inputObj) || inputObj is not string input)
+        {
+            throw new ArgumentException("MCP Tool invocation is missing required 'input' argument of type string.");
+        }
+
+        string workflowName = context.Name;
+        string orchestrationFunctionName = WorkflowNamingHelper.ToOrchestrationFunctionName(workflowName);
+
+        DurableWorkflowInput<string> orchestrationInput = new() { Input = input };
+        string instanceId = await client.ScheduleNewOrchestrationInstanceAsync(orchestrationFunctionName, orchestrationInput);
+
+        OrchestrationMetadata? metadata = await client.WaitForInstanceCompletionAsync(
+            instanceId,
+            getInputsAndOutputs: true,
+            cancellation: functionContext.CancellationToken);
+
+        if (metadata is null)
+        {
+            throw new InvalidOperationException($"Workflow orchestration '{instanceId}' returned no metadata.");
+        }
+
+        if (metadata.RuntimeStatus is OrchestrationRuntimeStatus.Failed)
+        {
+            string errorMessage = metadata.FailureDetails?.ErrorMessage ?? "Unknown error";
+            throw new InvalidOperationException($"Workflow orchestration '{instanceId}' failed: {errorMessage}");
+        }
+
+        if (metadata.RuntimeStatus is not OrchestrationRuntimeStatus.Completed)
+        {
+            throw new InvalidOperationException($"Workflow orchestration '{instanceId}' ended with unexpected status '{metadata.RuntimeStatus}'.");
+        }
+
+        return metadata.ReadOutputAs<DurableWorkflowResult>()?.Result;
+    }
+
    /// <summary>
    /// Creates an error response with the specified status code and error message.
    /// </summary>
@@ -2,6 +2,7 @@

 ## [Unreleased]

+- Added MCP tool trigger support for durable workflows ([#4768](https://github.com/microsoft/agent-framework/pull/4768))
 - Added Azure Functions hosting support for durable workflows ([#4436](https://github.com/microsoft/agent-framework/pull/4436))

 ## v1.0.0-preview.251219.1
@@ -6,7 +6,8 @@ namespace Microsoft.Agents.AI.Hosting.AzureFunctions;

 /// <summary>
 /// Provides access to agent-specific options for functions agents by name.
-/// Returns default options (HTTP trigger enabled, MCP tool disabled) when no explicit options were configured.
+/// Returns <see langword="false"/> when no explicit options have been configured for an agent,
+/// which distinguishes standalone agents from those auto-registered by workflows.
 /// </summary>
 internal sealed class DefaultFunctionsAgentOptionsProvider(IReadOnlyDictionary<string, FunctionsAgentOptions> functionsAgentOptions)
    : IFunctionsAgentOptionsProvider
@@ -14,32 +15,19 @@ internal sealed class DefaultFunctionsAgentOptionsProvider(IReadOnlyDictionary<s
    private readonly IReadOnlyDictionary<string, FunctionsAgentOptions> _functionsAgentOptions =
        functionsAgentOptions ?? throw new ArgumentNullException(nameof(functionsAgentOptions));

-    // Default options. HTTP trigger enabled, MCP tool disabled.
-    private static readonly FunctionsAgentOptions s_defaultOptions = new()
-    {
-        HttpTrigger = { IsEnabled = true },
-        McpToolTrigger = { IsEnabled = false }
-    };
-
    /// <summary>
    /// Attempts to retrieve the options associated with the specified agent name.
-    /// If not found, a default options instance (with HTTP trigger enabled) is returned.
+    /// Returns <see langword="false"/> when no options have been explicitly configured for the agent.
    /// </summary>
    /// <param name="agentName">The name of the agent whose options are to be retrieved. Cannot be null or empty.</param>
-    /// <param name="options">The options for the specified agent. Will never be null.</param>
-    /// <returns>Always true. Returns configured options if present; otherwise default fallback options.</returns>
+    /// <param name="options">
+    /// When this method returns <see langword="true"/>, contains the options for the specified agent;
+    /// otherwise, <see langword="null"/>.
+    /// </param>
+    /// <returns><see langword="true"/> if options were found for the agent; otherwise, <see langword="false"/>.</returns>
    public bool TryGet(string agentName, [NotNullWhen(true)] out FunctionsAgentOptions? options)
    {
        ArgumentException.ThrowIfNullOrEmpty(agentName);
-
-        if (this._functionsAgentOptions.TryGetValue(agentName, out FunctionsAgentOptions? existing))
-        {
-            options = existing;
-            return true;
-        }
-
-        // If not defined, return default options.
-        options = s_defaultOptions;
-        return true;
+        return this._functionsAgentOptions.TryGetValue(agentName, out options);
    }
 }
@@ -6,9 +6,13 @@ using Microsoft.Extensions.Logging;
 namespace Microsoft.Agents.AI.Hosting.AzureFunctions;

 /// <summary>
-/// Transforms function metadata by registering durable agent functions for each configured agent.
+/// Transforms function metadata by registering durable agent functions for each explicitly configured agent.
 /// </summary>
-/// <remarks>This transformer adds both entity trigger and HTTP trigger functions for every agent registered in the application.</remarks>
+/// <remarks>
+/// This transformer adds entity, HTTP, and MCP tool trigger functions for agents that have
+/// explicit <see cref="FunctionsAgentOptions"/>. Agents auto-registered by workflows
+/// (which lack explicit options) are handled by <see cref="DurableWorkflowsFunctionMetadataTransformer"/>.
+/// </remarks>
 internal sealed class DurableAgentFunctionMetadataTransformer : IFunctionMetadataTransformer
 {
    private readonly ILogger<DurableAgentFunctionMetadataTransformer> _logger;
@@ -38,24 +42,27 @@ internal sealed class DurableAgentFunctionMetadataTransformer : IFunctionMetadat
        {
            string agentName = kvp.Key;

-            this._logger.LogRegisteringTriggerForAgent(agentName, "entity");
+            // Only generate triggers for agents with explicit Functions agent options.
+            // Agents auto-registered by workflows are handled by DurableWorkflowsFunctionMetadataTransformer.
+            if (!this._functionsAgentOptionsProvider.TryGet(agentName, out FunctionsAgentOptions? agentTriggerOptions))
+            {
+                continue;
+            }

+            this._logger.LogRegisteringTriggerForAgent(agentName, "entity");
            original.Add(FunctionMetadataFactory.CreateEntityTrigger(agentName));

-            if (this._functionsAgentOptionsProvider.TryGet(agentName, out FunctionsAgentOptions? agentTriggerOptions))
+            if (agentTriggerOptions.HttpTrigger.IsEnabled)
            {
-                if (agentTriggerOptions.HttpTrigger.IsEnabled)
-                {
-                    this._logger.LogRegisteringTriggerForAgent(agentName, "http");
-                    original.Add(FunctionMetadataFactory.CreateHttpTrigger(agentName, $"agents/{agentName}/run", BuiltInFunctions.RunAgentHttpFunctionEntryPoint));
-                }
+                this._logger.LogRegisteringTriggerForAgent(agentName, "http");
+                original.Add(FunctionMetadataFactory.CreateHttpTrigger(agentName, $"agents/{agentName}/run", BuiltInFunctions.RunAgentHttpFunctionEntryPoint));
+            }

-                if (agentTriggerOptions.McpToolTrigger.IsEnabled)
-                {
-                    AIAgent agent = kvp.Value(this._serviceProvider);
-                    this._logger.LogRegisteringTriggerForAgent(agentName, "mcpTool");
-                    original.Add(CreateMcpToolTrigger(agentName, agent.Description));
-                }
+            if (agentTriggerOptions.McpToolTrigger.IsEnabled)
+            {
+                AIAgent agent = kvp.Value(this._serviceProvider);
+                this._logger.LogRegisteringTriggerForAgent(agentName, "mcpTool");
+                original.Add(CreateMcpToolTrigger(agentName, agent.Description));
            }
        }
    }
@@ -134,4 +134,17 @@ public static class DurableAgentsOptionsExtensions
    {
        return new Dictionary<string, FunctionsAgentOptions>(s_agentOptions, StringComparer.OrdinalIgnoreCase);
    }
+
+    /// <summary>
+    /// Ensures every agent in <paramref name="agentNames"/> has an entry in the
+    /// options registry. Agents that already have explicit options are left untouched.
+    /// New entries receive the default configuration (HTTP trigger enabled, MCP tool disabled).
+    /// </summary>
+    internal static void EnsureDefaultOptionsForAll(IEnumerable<string> agentNames)
+    {
+        foreach (string name in agentNames)
+        {
+            s_agentOptions.TryAdd(name, new FunctionsAgentOptions { HttpTrigger = { IsEnabled = true } });
+        }
+    }
 }
@@ -1,5 +1,6 @@
 // Copyright (c) Microsoft. All rights reserved.

+using System.Text.Json.Nodes;
 using Microsoft.Agents.AI.DurableTask;
 using Microsoft.Azure.Functions.Worker.Core.FunctionMetadata;

@@ -98,4 +99,65 @@ internal static class FunctionMetadataFactory
            ScriptFile = BuiltInFunctions.ScriptFile,
        };
    }
+
+    /// <summary>
+    /// Creates function metadata for an MCP tool trigger function that starts a workflow.
+    /// </summary>
+    /// <param name="workflowName">The name of the workflow to expose as an MCP tool.</param>
+    /// <param name="description">An optional description for the MCP tool. If null, a default description is generated.</param>
+    /// <returns>A <see cref="DefaultFunctionMetadata"/> configured for an MCP tool trigger.</returns>
+    internal static DefaultFunctionMetadata CreateWorkflowMcpToolTrigger(
+        string workflowName,
+        string? description)
+    {
+        var functionName = $"{BuiltInFunctions.McpToolPrefix}{workflowName}";
+        var toolDescription = description ?? $"Run the {workflowName} workflow";
+
+        var toolProperties = new JsonArray(new JsonObject
+        {
+            ["propertyName"] = "input",
+            ["propertyType"] = "string",
+            ["description"] = "The input to the workflow.",
+            ["isRequired"] = true,
+            ["isArray"] = false,
+        });
+
+        var triggerBinding = new JsonObject
+        {
+            ["name"] = "context",
+            ["type"] = "mcpToolTrigger",
+            ["direction"] = "In",
+            ["toolName"] = workflowName,
+            ["description"] = toolDescription,
+            ["toolProperties"] = toolProperties.ToJsonString(),
+        };
+
+        var inputBinding = new JsonObject
+        {
+            ["name"] = "input",
+            ["type"] = "mcpToolProperty",
+            ["direction"] = "In",
+            ["propertyName"] = "input",
+            ["description"] = "The input to the workflow",
+            ["isRequired"] = true,
+            ["dataType"] = "String",
+            ["propertyType"] = "string",
+        };
+
+        var clientBinding = new JsonObject
+        {
+            ["name"] = "client",
+            ["type"] = "durableClient",
+            ["direction"] = "In",
+        };
+
+        return new DefaultFunctionMetadata
+        {
+            Name = functionName,
+            Language = "dotnet-isolated",
+            RawBindings = [triggerBinding.ToJsonString(), inputBinding.ToJsonString(), clientBinding.ToJsonString()],
+            EntryPoint = BuiltInFunctions.RunWorkflowMcpToolFunctionEntryPoint,
+            ScriptFile = BuiltInFunctions.ScriptFile,
+        };
+    }
 }
@@ -27,9 +27,16 @@ public static class FunctionsApplicationBuilderExtensions
    {
        ArgumentNullException.ThrowIfNull(configure);

+        // Create/get shared options BEFORE the DurableTask library call so it can find them.
+        FunctionsDurableOptions sharedOptions = GetOrCreateSharedOptions(builder.Services);
+
        // The main agent services registration is done in Microsoft.DurableTask.Agents.
        builder.Services.ConfigureDurableAgents(configure);

+        // Ensure all agents registered through this path have default FunctionsAgentOptions.
+        // This distinguishes them from agents auto-registered by workflows.
+        DurableAgentsOptionsExtensions.EnsureDefaultOptionsForAll(sharedOptions.Agents.GetAgentFactories().Keys);
+
        builder.Services.TryAddSingleton<IFunctionsAgentOptionsProvider>(_ =>
            new DefaultFunctionsAgentOptionsProvider(DurableAgentsOptionsExtensions.GetAgentOptionsSnapshot()));

@@ -67,6 +74,13 @@ public static class FunctionsApplicationBuilderExtensions

        builder.Services.ConfigureDurableOptions(configure);

+        if (DurableAgentsOptionsExtensions.GetAgentOptionsSnapshot().Count > 0)
+        {
+            builder.Services.TryAddSingleton<IFunctionsAgentOptionsProvider>(_ =>
+                new DefaultFunctionsAgentOptionsProvider(DurableAgentsOptionsExtensions.GetAgentOptionsSnapshot()));
+            builder.Services.TryAddEnumerable(ServiceDescriptor.Singleton<IFunctionMetadataTransformer, DurableAgentFunctionMetadataTransformer>());
+        }
+
        if (sharedOptions.Workflows.Workflows.Count > 0)
        {
            builder.Services.TryAddEnumerable(ServiceDescriptor.Singleton<IFunctionMetadataTransformer, DurableWorkflowsFunctionMetadataTransformer>());
@@ -102,12 +116,14 @@ public static class FunctionsApplicationBuilderExtensions

        builder.UseWhen<BuiltInFunctionExecutionMiddleware>(static context =>
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunAgentHttpFunctionEntryPoint, StringComparison.Ordinal) ||
+            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunAgentMcpToolFunctionEntryPoint, StringComparison.Ordinal) ||
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunAgentEntityFunctionEntryPoint, StringComparison.Ordinal) ||
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunWorkflowOrchestrationHttpFunctionEntryPoint, StringComparison.Ordinal) ||
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunWorkflowOrchestrationFunctionEntryPoint, StringComparison.Ordinal) ||
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.InvokeWorkflowActivityFunctionEntryPoint, StringComparison.Ordinal) ||
            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.GetWorkflowStatusHttpFunctionEntryPoint, StringComparison.Ordinal) ||
-            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RespondToWorkflowHttpFunctionEntryPoint, StringComparison.Ordinal)
+            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RespondToWorkflowHttpFunctionEntryPoint, StringComparison.Ordinal) ||
+            string.Equals(context.FunctionDefinition.EntryPoint, BuiltInFunctions.RunWorkflowMcpToolFunctionEntryPoint, StringComparison.Ordinal)
        );
        builder.Services.TryAddSingleton<BuiltInFunctionExecutor>();
    }
@@ -10,6 +10,7 @@ namespace Microsoft.Agents.AI.Hosting.AzureFunctions;
 internal sealed class FunctionsDurableOptions : DurableOptions
 {
    private readonly HashSet<string> _statusEndpointWorkflows = new(StringComparer.OrdinalIgnoreCase);
+    private readonly HashSet<string> _mcpToolTriggerWorkflows = new(StringComparer.OrdinalIgnoreCase);

    /// <summary>
    /// Enables the status HTTP endpoint for the specified workflow.
@@ -26,4 +27,20 @@ internal sealed class FunctionsDurableOptions : DurableOptions
    {
        return this._statusEndpointWorkflows.Contains(workflowName);
    }
+
+    /// <summary>
+    /// Enables the MCP tool trigger for the specified workflow.
+    /// </summary>
+    internal void EnableMcpToolTrigger(string workflowName)
+    {
+        this._mcpToolTriggerWorkflows.Add(workflowName);
+    }
+
+    /// <summary>
+    /// Returns whether the MCP tool trigger is enabled for the specified workflow.
+    /// </summary>
+    internal bool IsMcpToolTriggerEnabled(string workflowName)
+    {
+        return this._mcpToolTriggerWorkflows.Contains(workflowName);
+    }
 }
@@ -27,4 +27,31 @@ public static class DurableWorkflowOptionsExtensions
            functionsOptions.EnableStatusEndpoint(workflow.Name!);
        }
    }
+
+    /// <summary>
+    /// Adds a workflow and configures whether to expose a status HTTP endpoint and/or an MCP tool trigger.
+    /// </summary>
+    /// <param name="options">The workflow options to add the workflow to.</param>
+    /// <param name="workflow">The workflow instance to add.</param>
+    /// <param name="exposeStatusEndpoint">If <see langword="true"/>, a GET endpoint is generated at <c>workflows/{name}/status/{runId}</c>.</param>
+    /// <param name="exposeMcpToolTrigger">If <see langword="true"/>, an MCP tool trigger is generated for the workflow.</param>
+    public static void AddWorkflow(this DurableWorkflowOptions options, Workflow workflow, bool exposeStatusEndpoint, bool exposeMcpToolTrigger)
+    {
+        ArgumentNullException.ThrowIfNull(options);
+
+        options.AddWorkflow(workflow);
+
+        if (options.ParentOptions is FunctionsDurableOptions functionsOptions)
+        {
+            if (exposeStatusEndpoint)
+            {
+                functionsOptions.EnableStatusEndpoint(workflow.Name!);
+            }
+
+            if (exposeMcpToolTrigger)
+            {
+                functionsOptions.EnableMcpToolTrigger(workflow.Name!);
+            }
+        }
+    }
 }
@@ -50,8 +50,11 @@ internal sealed class DurableWorkflowsFunctionMetadataTransformer : IFunctionMet
        int initialCount = original.Count;
        this._logger.LogTransformingFunctionMetadata(initialCount);

-        // Track registered function names to avoid duplicates when workflows share executors.
-        HashSet<string> registeredFunctions = [];
+        // Seed with existing function names to avoid duplicates across transformers
+        // (e.g., when DurableAgentFunctionMetadataTransformer already registered entity triggers).
+        HashSet<string> registeredFunctions = new(
+            original.Select(f => f.Name!),
+            StringComparer.OrdinalIgnoreCase);

        DurableWorkflowOptions workflowOptions = this._options.Workflows;
        foreach (var workflow in workflowOptions.Workflows)
@@ -113,6 +116,17 @@ internal sealed class DurableWorkflowsFunctionMetadataTransformer : IFunctionMet
                }
            }

+            // Register an MCP tool trigger if opted in via AddWorkflow(exposeMcpToolTrigger: true).
+            if (this._options.IsMcpToolTriggerEnabled(workflow.Key))
+            {
+                string mcpToolFunctionName = $"{BuiltInFunctions.McpToolPrefix}{workflow.Key}";
+                if (registeredFunctions.Add(mcpToolFunctionName))
+                {
+                    this._logger.LogRegisteringWorkflowTrigger(workflow.Key, mcpToolFunctionName, "mcpTool");
+                    original.Add(FunctionMetadataFactory.CreateWorkflowMcpToolTrigger(workflow.Key, workflow.Value.Description));
+                }
+            }
+
            // Register activity or entity functions for each executor in the workflow.
            // ReflectExecutors() returns all executors across the graph; no need to manually traverse edges.
            foreach (KeyValuePair<string, ExecutorBinding> entry in workflow.Value.ReflectExecutors())
@@ -38,6 +38,7 @@ internal static class SourceBuilder
        sb.AppendLine("using System.Collections.Generic;");
        sb.AppendLine("using Microsoft.Agents.AI.Workflows;");
        sb.AppendLine();
+        sb.AppendLine("using RouteBuilder = Microsoft.Agents.AI.Workflows.RouteBuilder;");

        // Namespace
        if (!string.IsNullOrWhiteSpace(info.Namespace))
@@ -5,11 +5,14 @@ using System.Collections.Generic;
 using System.IO;
 using System.Text;
 using System.Text.Json;
+using System.Text.Json.Serialization.Metadata;
 using System.Threading;
 using System.Threading.Tasks;

 namespace Microsoft.Agents.AI.Workflows.Checkpointing;

+internal record CheckpointFileIndexEntry(CheckpointInfo CheckpointInfo, string FileName);
+
 /// <summary>
 /// Provides a file system-based implementation of a JSON checkpoint store that persists checkpoint data and index
 /// information to disk using JSON files.
@@ -28,6 +31,8 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
    internal DirectoryInfo Directory { get; }
    internal HashSet<CheckpointInfo> CheckpointIndex { get; }

+    private static JsonTypeInfo<CheckpointFileIndexEntry> EntryTypeInfo => WorkflowsJsonUtilities.JsonContext.Default.CheckpointFileIndexEntry;
+
    /// <summary>
    /// Initializes a new instance of the <see cref="FileSystemJsonCheckpointStore"/> class that uses the specified directory
    /// </summary>
@@ -64,9 +69,11 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
            using StreamReader reader = new(this._indexFile, encoding: Encoding.UTF8, detectEncodingFromByteOrderMarks: false, BufferSize, leaveOpen: true);
            while (reader.ReadLine() is string line)
            {
-                if (JsonSerializer.Deserialize(line, KeyTypeInfo) is { } info)
+                if (JsonSerializer.Deserialize(line, EntryTypeInfo) is { } entry)
                {
-                    this.CheckpointIndex.Add(info);
+                    // We never actually use the file names from the index entries since they can be derived from the CheckpointInfo, but it is useful to
+                    // have the UrlEncoded file names in the index file for human readability
+                    this.CheckpointIndex.Add(entry.CheckpointInfo);
                }
            }
        }
@@ -93,8 +100,14 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
        }
    }

-    private string GetFileNameForCheckpoint(string sessionId, CheckpointInfo key)
-        => Path.Combine(this.Directory.FullName, $"{sessionId}_{key.CheckpointId}.json");
+    internal string GetFileNameForCheckpoint(string sessionId, CheckpointInfo key)
+    {
+        string protoPath = $"{sessionId}_{key.CheckpointId}.json";
+
+        // Escape the protoPath to ensure it is a valid file name, especially if sessionId or CheckpointId contain path separators, etc.
+        return Uri.EscapeDataString(protoPath) // This takes care of most of the invalid path characters
+                  .Replace(".", "%2E");        // This takes care of escaping the root folder, since EscapeDataString does not escape dots
+    }

    private CheckpointInfo GetUnusedCheckpointInfo(string sessionId)
    {
@@ -116,13 +129,16 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos

        CheckpointInfo key = this.GetUnusedCheckpointInfo(sessionId);
        string fileName = this.GetFileNameForCheckpoint(sessionId, key);
+        string filePath = Path.Combine(this.Directory.FullName, fileName);
+
        try
        {
-            using Stream checkpointStream = File.Open(fileName, FileMode.Create, FileAccess.Write, FileShare.None);
+            using Stream checkpointStream = File.Open(filePath, FileMode.Create, FileAccess.Write, FileShare.None);
            using Utf8JsonWriter jsonWriter = new(checkpointStream, new JsonWriterOptions() { Indented = false });
            value.WriteTo(jsonWriter);

-            JsonSerializer.Serialize(this._indexFile!, key, KeyTypeInfo);
+            CheckpointFileIndexEntry entry = new(key, fileName);
+            JsonSerializer.Serialize(this._indexFile!, entry, EntryTypeInfo);
            byte[] bytes = Encoding.UTF8.GetBytes(Environment.NewLine);
            await this._indexFile!.WriteAsync(bytes, 0, bytes.Length, CancellationToken.None).ConfigureAwait(false);
            await this._indexFile!.FlushAsync(CancellationToken.None).ConfigureAwait(false);
@@ -136,7 +152,7 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
            try
            {
                // try to clean up after ourselves
-                File.Delete(fileName);
+                File.Delete(filePath);
            }
            catch { }

@@ -149,6 +165,7 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
    {
        this.CheckDisposed();
        string fileName = this.GetFileNameForCheckpoint(sessionId, key);
+        string filePath = Path.Combine(this.Directory.FullName, fileName);

        if (!this.CheckpointIndex.Contains(key) ||
            !File.Exists(fileName))
@@ -156,7 +173,7 @@ public sealed class FileSystemJsonCheckpointStore : JsonCheckpointStore, IDispos
            throw new KeyNotFoundException($"Checkpoint '{key.CheckpointId}' not found in store at '{this.Directory.FullName}'.");
        }

-        using FileStream checkpointFileStream = File.Open(fileName, FileMode.Open, FileAccess.Read, FileShare.Read);
+        using FileStream checkpointFileStream = File.Open(filePath, FileMode.Open, FileAccess.Read, FileShare.Read);
        using JsonDocument document = await JsonDocument.ParseAsync(checkpointFileStream).ConfigureAwait(false);

        return document.RootElement.Clone();
@@ -3,9 +3,9 @@
 namespace Microsoft.Agents.AI.Workflows;

 /// <summary>
-/// Provides extensions methods for creating <see cref="Configured{TSubject}"/> objects
+/// Provides extension methods for creating <see cref="Configured{TSubject}"/> objects
 /// </summary>
-public static class ConfigurationExtensions
+internal static class ConfigurationExtensions
 {
    /// <summary>
    /// Creates a new configuration that treats the subject as its base type, allowing configuration to be applied at
@@ -8,7 +8,7 @@ namespace Microsoft.Agents.AI.Workflows;
 /// <summary>
 /// Provides methods for creating <see cref="Configured{TSubject}"/> instances.
 /// </summary>
-public static class Configured
+internal static class Configured
 {
    /// <summary>
    /// Creates a <see cref="Configured{TSubject}"/> instance from an existing subject instance.
@@ -50,10 +50,10 @@ public static class Configured
 /// A representation of a preconfigured, lazy-instantiatable instance of <typeparamref name="TSubject"/>.
 /// </summary>
 /// <typeparam name="TSubject">The type of the preconfigured subject.</typeparam>
-/// <param name="factoryAsync">A factory to intantiate the subject when desired.</param>
+/// <param name="factoryAsync">A factory to instantiate the subject when desired.</param>
 /// <param name="id">The unique identifier for the configured subject.</param>
 /// <param name="raw"></param>
-public class Configured<TSubject>(Func<Config, string, ValueTask<TSubject>> factoryAsync, string id, object? raw = null)
+internal class Configured<TSubject>(Func<ExecutorConfig, string, ValueTask<TSubject>> factoryAsync, string id, object? raw = null)
 {
    /// <summary>
    /// Gets the raw representation of the configured object, if any.
@@ -66,14 +66,14 @@ public class Configured<TSubject>(Func<Config, string, ValueTask<TSubject>> fact
    public string Id => id;

    /// <summary>
-    /// Gets the factory function to create an instance of <typeparamref name="TSubject"/> given a <see cref="Config"/>.
+    /// Gets the factory function to create an instance of <typeparamref name="TSubject"/> given a <see cref="ExecutorConfig"/>.
    /// </summary>
-    public Func<Config, string, ValueTask<TSubject>> FactoryAsync => factoryAsync;
+    public Func<ExecutorConfig, string, ValueTask<TSubject>> FactoryAsync => factoryAsync;

    /// <summary>
    /// The configuration for this configured instance.
    /// </summary>
-    public Config Configuration => new(this.Id);
+    public ExecutorConfig Configuration => new(this.Id);

    /// <summary>
    /// Gets a "partially" applied factory function that only requires no parameters to create an instance of
@@ -87,11 +87,11 @@ public class Configured<TSubject>(Func<Config, string, ValueTask<TSubject>> fact
 /// </summary>
 /// <typeparam name="TSubject">The type of the preconfigured subject.</typeparam>
 /// <typeparam name="TOptions">The type of configuration options for the preconfigured subject.</typeparam>
-/// <param name="factoryAsync">A factory to intantiate the subject when desired.</param>
+/// <param name="factoryAsync">A factory to instantiate the subject when desired.</param>
 /// <param name="id">The unique identifier for the configured subject.</param>
 /// <param name="options">Additional configuration options for the subject.</param>
 /// <param name="raw"></param>
-public class Configured<TSubject, TOptions>(Func<Config<TOptions>, string, ValueTask<TSubject>> factoryAsync, string id, TOptions? options = default, object? raw = null)
+internal class Configured<TSubject, TOptions>(Func<ExecutorConfig<TOptions>, string, ValueTask<TSubject>> factoryAsync, string id, TOptions? options = default, object? raw = null)
 {
    /// <summary>
    /// The raw representation of the configured object, if any.
@@ -109,14 +109,14 @@ public class Configured<TSubject, TOptions>(Func<Config<TOptions>, string, Value
    public TOptions? Options => options;

    /// <summary>
-    /// Gets the factory function to create an instance of <typeparamref name="TSubject"/> given a <see cref="Config{TOptions}"/>.
+    /// Gets the factory function to create an instance of <typeparamref name="TSubject"/> given a <see cref="ExecutorConfig{TOptions}"/>.
    /// </summary>
-    public Func<Config<TOptions>, string, ValueTask<TSubject>> FactoryAsync => factoryAsync;
+    public Func<ExecutorConfig<TOptions>, string, ValueTask<TSubject>> FactoryAsync => factoryAsync;

    /// <summary>
    /// The configuration for this configured instance.
    /// </summary>
-    public Config<TOptions> Configuration => new(this.Id, this.Options);
+    public ExecutorConfig<TOptions> Configuration => new(this.Id, this.Options);

    /// <summary>
    /// Gets a "partially" applied factory function that only requires no parameters to create an instance of
@@ -124,11 +124,11 @@ public class Configured<TSubject, TOptions>(Func<Config<TOptions>, string, Value
    /// </summary>
    internal Func<string, ValueTask<TSubject>> BoundFactoryAsync => (sessionId) => this.CreateValidatingMemoizedFactory()(this.Configuration, sessionId);

-    private Func<Config, string, ValueTask<TSubject>> CreateValidatingMemoizedFactory()
+    private Func<ExecutorConfig, string, ValueTask<TSubject>> CreateValidatingMemoizedFactory()
    {
        return FactoryAsync;

-        async ValueTask<TSubject> FactoryAsync(Config configuration, string sessionId)
+        async ValueTask<TSubject> FactoryAsync(ExecutorConfig configuration, string sessionId)
        {
            if (this.Id != configuration.Id)
            {
@@ -53,6 +53,9 @@ internal sealed class AsyncRunHandle : ICheckpointingHandle, IAsyncDisposable
    public ValueTask<RunStatus> GetStatusAsync(CancellationToken cancellationToken = default)
        => this._eventStream.GetStatusAsync(cancellationToken);

+    internal bool TryGetResponsePortExecutorId(string portId, out string? executorId)
+        => this._stepRunner.TryGetResponsePortExecutorId(portId, out executorId);
+
    public async IAsyncEnumerable<WorkflowEvent> TakeEventStreamAsync(bool blockOnPendingRequest, [EnumeratorCancellation] CancellationToken cancellationToken = default)
    {
        //Debug.Assert(breakOnHalt);
@@ -3,6 +3,7 @@
 using System;
 using System.Collections.Concurrent;
 using System.Collections.Generic;
+using System.Diagnostics.CodeAnalysis;
 using System.Linq;
 using System.Threading;
 using System.Threading.Tasks;
@@ -95,6 +96,18 @@ internal sealed class EdgeMap
        return portRunner.ChaseEdgeAsync(new MessageEnvelope(response, ExecutorIdentity.None), this._stepTracer, cancellationToken);
    }

+    internal bool TryGetResponsePortExecutorId(string portId, [NotNullWhen(true)] out string? executorId)
+    {
+        if (this._portEdgeRunners.TryGetValue(portId, out ResponseEdgeRunner? portRunner))
+        {
+            executorId = portRunner.ExecutorId;
+            return true;
+        }
+
+        executorId = null;
+        return false;
+    }
+
    internal async ValueTask<Dictionary<EdgeId, PortableValue>> ExportStateAsync()
    {
        Dictionary<EdgeId, PortableValue> exportedStates = [];
@@ -19,6 +19,7 @@ internal interface ISuperStepRunner
    bool HasUnprocessedMessages { get; }

    ValueTask EnqueueResponseAsync(ExternalResponse response, CancellationToken cancellationToken = default);
+    bool TryGetResponsePortExecutorId(string portId, out string? executorId);

    ValueTask<bool> IsValidInputTypeAsync<T>(CancellationToken cancellationToken = default);
    ValueTask<bool> EnqueueMessageAsync<T>(T message, CancellationToken cancellationToken = default);
@@ -106,8 +106,7 @@ internal sealed class StateManager
        if (typeof(T) == typeof(object))
        {
            // Reading as object will break across serialize/deserialize boundaries, e.g. checkpointing, distributed runtime, etc.
-            // Disabled pending upstream updates for this change; see https://github.com/microsoft/agent-framework/issues/1369
-            //throw new NotSupportedException("Reading state as 'object' is not supported. Use 'PortableValue' instead for variants.");
+            throw new NotSupportedException("Reading state as 'object' is not supported. Use 'PortableValue' instead for variants.");
        }

        Throw.IfNullOrEmpty(key);
@@ -113,7 +113,7 @@ public static class ExecutorBindingExtensions
    /// <param name="id">An id for the executor to be instantiated.</param>
    /// <param name="options">An optional parameter specifying the options.</param>
    /// <returns>An <see cref="ExecutorBinding"/> instance that resolves to the result of the factory call when messages get sent to it.</returns>
-    public static ExecutorBinding BindExecutor<TExecutor, TOptions>(this Func<Config<TOptions>, string, ValueTask<TExecutor>> factoryAsync, string id, TOptions? options = null)
+    public static ExecutorBinding BindExecutor<TExecutor, TOptions>(this Func<ExecutorConfig<TOptions>, string, ValueTask<TExecutor>> factoryAsync, string id, TOptions? options = null)
        where TExecutor : Executor
        where TOptions : ExecutorOptions
    {
@@ -139,7 +139,7 @@ public static class ExecutorBindingExtensions
    /// <returns>An <see cref="ExecutorBinding"/> instance that resolves to the result of the factory call when messages get sent to it.</returns>
    [Obsolete("Use BindExecutor() instead")]
    [EditorBrowsable(EditorBrowsableState.Never)]
-    public static ExecutorBinding ConfigureFactory<TExecutor, TOptions>(this Func<Config<TOptions>, string, ValueTask<TExecutor>> factoryAsync, string id, TOptions? options = null)
+    public static ExecutorBinding ConfigureFactory<TExecutor, TOptions>(this Func<ExecutorConfig<TOptions>, string, ValueTask<TExecutor>> factoryAsync, string id, TOptions? options = null)
        where TExecutor : Executor
        where TOptions : ExecutorOptions
        => factoryAsync.BindExecutor(id, options);
@@ -6,7 +6,7 @@ namespace Microsoft.Agents.AI.Workflows;
 /// Represents a configuration for an object with a string identifier. For example, <see cref="IIdentified"/> object.
 /// </summary>
 /// <param name="id">A unique identifier for the configurable object.</param>
-public class Config(string id)
+public class ExecutorConfig(string id)
 {
    /// <summary>
    /// Gets a unique identifier for the configurable object.
@@ -23,7 +23,7 @@ public class Config(string id)
 /// <typeparam name="TOptions">The type of options for the configurable object.</typeparam>
 /// <param name="id">A unique identifier for the configurable object.</param>
 /// <param name="options">The options for the configurable object.</param>
-public class Config<TOptions>(string id, TOptions? options = default) : Config(id)
+public class ExecutorConfig<TOptions>(string id, TOptions? options = default) : ExecutorConfig(id)
 {
    /// <summary>
    /// Gets the options for the configured object.
@@ -13,10 +13,18 @@ namespace Microsoft.Agents.AI.Workflows;
 /// </summary>
 public sealed class HandoffsWorkflowBuilder
 {
-    internal const string FunctionPrefix = "handoff_to_";
+    /// <summary>
+    /// The prefix for function calls that trigger handoffs to other agents; the full name is then `{FunctionPrefix}&lt;agent_id&gt;`,
+    /// where `&lt;agent_id&gt;` is the ID of the target agent to hand off to.
+    /// </summary>
+    public const string FunctionPrefix = "handoff_to_";
+
    private readonly AIAgent _initialAgent;
    private readonly Dictionary<AIAgent, HashSet<HandoffTarget>> _targets = [];
    private readonly HashSet<AIAgent> _allAgents = new(AIAgentIDEqualityComparer.Instance);
+
+    private bool _emitAgentResponseEvents;
+    private bool _emitAgentResponseUpdateEvents;
    private HandoffToolCallFilteringBehavior _toolCallFilteringBehavior = HandoffToolCallFilteringBehavior.HandoffOnly;

    /// <summary>
@@ -47,9 +55,13 @@ public sealed class HandoffsWorkflowBuilder
              """;

    /// <summary>
-    /// Sets additional instructions to provide to an agent that has handoffs about how and when to
-    /// perform them.
+    /// Sets instructions to provide to each agent that has handoffs about how and when to perform them.
    /// </summary>
+    /// <remarks>
+    /// In the vast majority of cases, the <see cref="DefaultHandoffInstructions"/> will be sufficient, and there will be no need to customize.
+    /// If you do provide alternate instructions, remember to explain the mechanics of the handoff function tool call, using see
+    /// <see cref="FunctionPrefix"/> constant.
+    /// </remarks>
    /// <param name="instructions">The instructions to provide, or <see langword="null"/> to restore the default instructions.</param>
    public HandoffsWorkflowBuilder WithHandoffInstructions(string? instructions)
    {
@@ -57,6 +69,29 @@ public sealed class HandoffsWorkflowBuilder
        return this;
    }

+    /// <summary>
+    /// Sets a value indicating whether agent streaming update events should be emitted during execution.
+    /// If <see langword="null"/>, the value will be taken from the <see cref="TurnToken"/>
+    /// </summary>
+    /// <param name="emitAgentResponseUpdateEvents"></param>
+    /// <returns></returns>
+    public HandoffsWorkflowBuilder EmitAgentResponseUpdateEvents(bool emitAgentResponseUpdateEvents = true)
+    {
+        this._emitAgentResponseUpdateEvents = emitAgentResponseUpdateEvents;
+        return this;
+    }
+
+    /// <summary>
+    /// Sets a value indicating whether aggregated agent response events should be emitted during execution.
+    /// </summary>
+    /// <param name="emitAgentResponseEvents"></param>
+    /// <returns></returns>
+    public HandoffsWorkflowBuilder EmitAgentResponseEvents(bool emitAgentResponseEvents = true)
+    {
+        this._emitAgentResponseEvents = emitAgentResponseEvents;
+        return this;
+    }
+
    /// <summary>
    /// Sets the behavior for filtering <see cref="FunctionCallContent"/> and <see cref="ChatRole.Tool"/> contents from
    /// <see cref="ChatMessage"/>s flowing through the handoff workflow. Defaults to <see cref="HandoffToolCallFilteringBehavior.HandoffOnly"/>.
@@ -175,7 +210,10 @@ public sealed class HandoffsWorkflowBuilder
        HandoffsEndExecutor end = new();
        WorkflowBuilder builder = new(start);

-        HandoffAgentExecutorOptions options = new(this.HandoffInstructions, this._toolCallFilteringBehavior);
+        HandoffAgentExecutorOptions options = new(this.HandoffInstructions,
+                                                  this._emitAgentResponseEvents,
+                                                  this._emitAgentResponseUpdateEvents,
+                                                  this._toolCallFilteringBehavior);

        // Create an AgentExecutor for each again.
        Dictionary<string, HandoffAgentExecutor> executors = this._allAgents.ToDictionary(a => a.Id, a => new HandoffAgentExecutor(a, options));
@@ -160,6 +160,8 @@ internal sealed class InProcessRunner : ISuperStepRunner, ICheckpointingHandle

    bool ISuperStepRunner.HasUnservicedRequests => this.RunContext.HasUnservicedRequests;
    bool ISuperStepRunner.HasUnprocessedMessages => this.RunContext.NextStepHasActions;
+    bool ISuperStepRunner.TryGetResponsePortExecutorId(string portId, out string? executorId)
+        => this.RunContext.TryGetResponsePortExecutorId(portId, out executorId);

    public bool IsCheckpointingEnabled => this.RunContext.IsCheckpointingEnabled;

@@ -296,6 +296,9 @@ internal sealed class InProcessRunnerContext : IRunnerContext
        return this._externalRequests.TryRemove(requestId, out _);
    }

+    internal bool TryGetResponsePortExecutorId(string portId, [NotNullWhen(true)] out string? executorId)
+        => this._edgeMap.TryGetResponsePortExecutorId(portId, out executorId);
+
    private IEventSink OutgoingEvents { get; }

    internal StateManager StateManager { get; } = new();
@@ -12,6 +12,15 @@ namespace Microsoft.Agents.AI.Workflows.Specialized;

 internal record AIAgentHostState(JsonElement? ThreadState, bool? CurrentTurnEmitEvents);

+internal static class TurnExtensions
+{
+    public static bool ShouldEmitStreamingEvents(this TurnToken token, bool? agentSetting)
+        => token.EmitEvents ?? agentSetting ?? false;
+
+    public static bool ShouldEmitStreamingEvents(bool? turnTokenSetting, bool? agentSetting)
+        => turnTokenSetting ?? agentSetting ?? false;
+}
+
 internal sealed class AIAgentHostExecutor : ChatProtocolExecutor
 {
    private readonly AIAgent _agent;
@@ -68,10 +77,17 @@ internal sealed class AIAgentHostExecutor : ChatProtocolExecutor
            throw new InvalidOperationException($"No pending ToolApprovalRequest found with id '{response.RequestId}'.");
        }

-        List<ChatMessage> implicitTurnMessages = [new ChatMessage(ChatRole.User, [response])];
+        // Merge the external response with any already-buffered regular messages so mixed-content
+        // resumes can be processed in one invocation.
+        return this.ProcessTurnMessagesAsync(async (pendingMessages, ctx, ct) =>
+        {
+            pendingMessages.Add(new ChatMessage(ChatRole.User, [response]));

-        // ContinueTurnAsync owns failing to emit a TurnToken if this response does not clear up all remaining outstanding requests.
-        return this.ContinueTurnAsync(implicitTurnMessages, context, this._currentTurnEmitEvents ?? false, cancellationToken);
+            await this.ContinueTurnAsync(pendingMessages, ctx, this._currentTurnEmitEvents ?? false, ct).ConfigureAwait(false);
+
+            // Clear the buffered turn messages because they were consumed by ContinueTurnAsync.
+            return null;
+        }, context, cancellationToken);
    }

    private ValueTask HandleFunctionResultAsync(
@@ -84,12 +100,18 @@ internal sealed class AIAgentHostExecutor : ChatProtocolExecutor
            throw new InvalidOperationException($"No pending FunctionCall found with id '{result.CallId}'.");
        }

-        List<ChatMessage> implicitTurnMessages = [new ChatMessage(ChatRole.Tool, [result])];
-        return this.ContinueTurnAsync(implicitTurnMessages, context, this._currentTurnEmitEvents ?? false, cancellationToken);
-    }
+        // Merge the external response with any already-buffered regular messages so mixed-content
+        // resumes can be processed in one invocation.
+        return this.ProcessTurnMessagesAsync(async (pendingMessages, ctx, ct) =>
+        {
+            pendingMessages.Add(new ChatMessage(ChatRole.Tool, [result]));

-    public bool ShouldEmitStreamingEvents(bool? emitEvents)
-        => emitEvents ?? this._options.EmitAgentUpdateEvents ?? false;
+            await this.ContinueTurnAsync(pendingMessages, ctx, this._currentTurnEmitEvents ?? false, ct).ConfigureAwait(false);
+
+            // Clear the buffered turn messages because they were consumed by ContinueTurnAsync.
+            return null;
+        }, context, cancellationToken);
+    }

    private async ValueTask<AgentSession> EnsureSessionAsync(IWorkflowContext context, CancellationToken cancellationToken) =>
        this._session ??= await this._agent.CreateSessionAsync(cancellationToken).ConfigureAwait(false);
@@ -159,7 +181,10 @@ internal sealed class AIAgentHostExecutor : ChatProtocolExecutor
    }

    protected override ValueTask TakeTurnAsync(List<ChatMessage> messages, IWorkflowContext context, bool? emitEvents, CancellationToken cancellationToken = default)
-        => this.ContinueTurnAsync(messages, context, this.ShouldEmitStreamingEvents(emitEvents), cancellationToken);
+        => this.ContinueTurnAsync(messages,
+                                  context,
+                                  TurnExtensions.ShouldEmitStreamingEvents(turnTokenSetting: emitEvents, this._options.EmitAgentUpdateEvents),
+                                  cancellationToken);

    private async ValueTask<AgentResponse> InvokeAgentAsync(IEnumerable<ChatMessage> messages, IWorkflowContext context, bool emitEvents, CancellationToken cancellationToken = default)
    {
@@ -198,7 +223,7 @@ internal sealed class AIAgentHostExecutor : ChatProtocolExecutor
            ExtractUnservicedRequests(response.Messages.SelectMany(message => message.Contents));
        }

-        if (this._options.EmitAgentResponseEvents == true)
+        if (this._options.EmitAgentResponseEvents)
        {
            await context.YieldOutputAsync(response, cancellationToken).ConfigureAwait(false);
        }
@@ -16,10 +16,12 @@ internal sealed class AIContentExternalHandler<TRequestContent, TResponseContent
    where TResponseContent : AIContent
 {
    private readonly PortBinding? _portBinding;
+    private readonly string _portId;
    private ConcurrentDictionary<string, TRequestContent> _pendingRequests = new();

    public AIContentExternalHandler(ref ProtocolBuilder protocolBuilder, string portId, bool intercepted, Func<TResponseContent, IWorkflowContext, CancellationToken, ValueTask> handler)
    {
+        this._portId = portId;
        PortBinding? portBinding = null;
        protocolBuilder = protocolBuilder.ConfigureRoutes(routeBuilder => ConfigureRoutes(routeBuilder, out portBinding));
        this._portBinding = portBinding;
@@ -58,12 +60,14 @@ internal sealed class AIContentExternalHandler<TRequestContent, TResponseContent
    {
        if (!this._pendingRequests.TryAdd(id, requestContent))
        {
-            throw new InvalidOperationException($"A pending request with ID '{id}' already exists.");
+            // Request is already pending; treat as an idempotent re-emission.
+            // Do not repost to the sink because request IDs must remain unique while pending.
+            return default;
        }

        return this.IsIntercepted
             ? context.SendMessageAsync(requestContent, cancellationToken: cancellationToken)
-             : this._portBinding.PostRequestAsync(requestContent, id, cancellationToken);
+             : this._portBinding.PostRequestAsync(requestContent, this.CreateExternalRequestId(id), cancellationToken);
    }

    public bool MarkRequestAsHandled(string id)
@@ -74,6 +78,8 @@ internal sealed class AIContentExternalHandler<TRequestContent, TResponseContent
    [MemberNotNullWhen(false, nameof(_portBinding))]
    private bool IsIntercepted => this._portBinding == null;

+    private string CreateExternalRequestId(string requestId) => $"{this._portId.Length}:{this._portId}:{requestId}";
+
    private static string MakeKey(string id) => $"{id}_PendingRequests";

    public async ValueTask OnCheckpointingAsync(string id, IWorkflowContext context, CancellationToken cancellationToken = default)
@@ -14,14 +14,20 @@ namespace Microsoft.Agents.AI.Workflows.Specialized;

 internal sealed class HandoffAgentExecutorOptions
 {
-    public HandoffAgentExecutorOptions(string? handoffInstructions, HandoffToolCallFilteringBehavior toolCallFilteringBehavior)
+    public HandoffAgentExecutorOptions(string? handoffInstructions, bool emitAgentResponseEvents, bool? emitAgentResponseUpdateEvents, HandoffToolCallFilteringBehavior toolCallFilteringBehavior)
    {
        this.HandoffInstructions = handoffInstructions;
+        this.EmitAgentResponseEvents = emitAgentResponseEvents;
+        this.EmitAgentResponseUpdateEvents = emitAgentResponseUpdateEvents;
        this.ToolCallFilteringBehavior = toolCallFilteringBehavior;
    }

    public string? HandoffInstructions { get; set; }

+    public bool EmitAgentResponseEvents { get; set; }
+
+    public bool? EmitAgentResponseUpdateEvents { get; set; }
+
    public HandoffToolCallFilteringBehavior ToolCallFilteringBehavior { get; set; } = HandoffToolCallFilteringBehavior.HandoffOnly;
 }

@@ -250,7 +256,14 @@ internal sealed class HandoffAgentExecutor(
            }
        }

-        allMessages.AddRange(updates.ToAgentResponse().Messages);
+        AgentResponse agentResponse = updates.ToAgentResponse();
+
+        if (options.EmitAgentResponseEvents)
+        {
+            await context.YieldOutputAsync(agentResponse, cancellationToken).ConfigureAwait(false);
+        }
+
+        allMessages.AddRange(agentResponse.Messages);

        roleChanges.ResetUserToAssistantForChangedRoles();

@@ -259,7 +272,7 @@ internal sealed class HandoffAgentExecutor(
        async Task AddUpdateAsync(AgentResponseUpdate update, CancellationToken cancellationToken)
        {
            updates.Add(update);
-            if (message.TurnToken.EmitEvents is true)
+            if (message.TurnToken.ShouldEmitStreamingEvents(options.EmitAgentResponseUpdateEvents))
            {
                await context.YieldOutputAsync(update, cancellationToken).ConfigureAwait(false);
            }
@@ -60,6 +60,9 @@ public sealed class StreamingRun : CheckpointableRunBase, IAsyncDisposable
    internal ValueTask<bool> TrySendMessageUntypedAsync(object message, Type? declaredType = null)
        => this._runHandle.EnqueueMessageUntypedAsync(message, declaredType);

+    internal bool TryGetResponsePortExecutorId(string portId, out string? executorId)
+        => this._runHandle.TryGetResponsePortExecutorId(portId, out executorId);
+
    /// <summary>
    /// Asynchronously streams workflow events as they occur during workflow execution.
    /// </summary>
@@ -43,7 +43,7 @@ internal sealed class WorkflowChatHistoryProvider : ChatHistoryProvider
        => this._sessionState.GetOrInitializeState(session).Messages.AddRange(messages);

    protected override ValueTask<IEnumerable<ChatMessage>> ProvideChatHistoryAsync(InvokingContext context, CancellationToken cancellationToken = default)
-        => new(this._sessionState.GetOrInitializeState(context.Session).Messages);
+        => new(this._sessionState.GetOrInitializeState(context.Session).Messages.AsReadOnly());

    protected override ValueTask StoreChatHistoryAsync(InvokedContext context, CancellationToken cancellationToken = default)
    {
@@ -62,6 +62,12 @@ internal sealed class WorkflowChatHistoryProvider : ChatHistoryProvider
        }
    }

+    public IEnumerable<ChatMessage> GetAllMessages(AgentSession session)
+    {
+        var state = this._sessionState.GetOrInitializeState(session);
+        return state.Messages.AsReadOnly();
+    }
+
    public void UpdateBookmark(AgentSession session)
    {
        var state = this._sessionState.GetOrInitializeState(session);
@@ -119,13 +119,17 @@ internal sealed class WorkflowHostAgent : AIAgent
        MessageMerger merger = new();

        await foreach (AgentResponseUpdate update in workflowSession.InvokeStageAsync(cancellationToken)
-                                                                      .ConfigureAwait(false)
-                                                                      .WithCancellation(cancellationToken))
+                                                                     .ConfigureAwait(false)
+                                                                     .WithCancellation(cancellationToken))
        {
            merger.AddUpdate(update);
        }

-        return merger.ComputeMerged(workflowSession.LastResponseId!, this.Id, this.Name);
+        AgentResponse response = merger.ComputeMerged(workflowSession.LastResponseId!, this.Id, this.Name);
+        workflowSession.ChatHistoryProvider.AddMessages(workflowSession, response.Messages);
+        workflowSession.ChatHistoryProvider.UpdateBookmark(workflowSession);
+
+        return response;
    }

    protected override async
@@ -138,11 +142,18 @@ internal sealed class WorkflowHostAgent : AIAgent
        await this.ValidateWorkflowAsync().ConfigureAwait(false);

        WorkflowSession workflowSession = await this.UpdateSessionAsync(messages, session, cancellationToken).ConfigureAwait(false);
+        MessageMerger merger = new();
+
        await foreach (AgentResponseUpdate update in workflowSession.InvokeStageAsync(cancellationToken)
                                                                      .ConfigureAwait(false)
                                                                      .WithCancellation(cancellationToken))
        {
+            merger.AddUpdate(update);
            yield return update;
        }
+
+        AgentResponse response = merger.ComputeMerged(workflowSession.LastResponseId!, this.Id, this.Name);
+        workflowSession.ChatHistoryProvider.AddMessages(workflowSession, response.Messages);
+        workflowSession.ChatHistoryProvider.UpdateBookmark(workflowSession);
    }
 }
@@ -25,6 +25,25 @@ internal sealed class WorkflowSession : AgentSession

    private InMemoryCheckpointManager? _inMemoryCheckpointManager;

+    /// <summary>
+    /// Tracks pending external requests by their workflow-facing request ID.
+    /// This mapping enables converting incoming response content back to <see cref="ExternalResponse"/>
+    /// when resuming a workflow from a checkpoint.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// Entries are added when a <see cref="RequestInfoEvent"/> is received during workflow execution,
+    /// and removed when a matching response is delivered via <see cref="SendMessagesWithResponseConversionAsync"/>.
+    /// </para>
+    /// <para>
+    /// The number of entries is bounded by the number of outstanding external requests in a single workflow run.
+    /// When a session is abandoned, all pending requests are released with the session object.
+    /// Request-level timeouts, if needed, should be implemented in the workflow definition itself
+    /// (e.g., using a timer racing against an external event).
+    /// </para>
+    /// </remarks>
+    private readonly Dictionary<string, ExternalRequest> _pendingRequests = [];
+
    internal static bool VerifyCheckpointingConfiguration(IWorkflowExecutionEnvironment executionEnvironment, [NotNullWhen(true)] out InProcessExecutionEnvironment? inProcEnv)
    {
        inProcEnv = null;
@@ -90,6 +109,7 @@ internal sealed class WorkflowSession : AgentSession

        this.LastCheckpoint = sessionState.LastCheckpoint;
        this.StateBag = sessionState.StateBag;
+        this._pendingRequests = sessionState.PendingRequests ?? [];
    }

    public CheckpointInfo? LastCheckpoint { get; set; }
@@ -101,7 +121,8 @@ internal sealed class WorkflowSession : AgentSession
            this.SessionId,
            this.LastCheckpoint,
            this._inMemoryCheckpointManager,
-            this.StateBag);
+            this.StateBag,
+            this._pendingRequests);

        return marshaller.Marshal(info);
    }
@@ -110,7 +131,7 @@ internal sealed class WorkflowSession : AgentSession
    {
        Throw.IfNullOrEmpty(parts);

-        AgentResponseUpdate update = new(ChatRole.Assistant, parts)
+        return new(ChatRole.Assistant, parts)
        {
            CreatedAt = DateTimeOffset.UtcNow,
            MessageId = Guid.NewGuid().ToString("N"),
@@ -118,30 +139,22 @@ internal sealed class WorkflowSession : AgentSession
            ResponseId = responseId,
            RawRepresentation = raw
        };
-
-        this.ChatHistoryProvider.AddMessages(this, update.ToChatMessage());
-
-        return update;
    }

    public AgentResponseUpdate CreateUpdate(string responseId, object raw, ChatMessage message)
    {
        Throw.IfNull(message);

-        AgentResponseUpdate update = new(message.Role, message.Contents)
+        return new(message.Role, message.Contents)
        {
            CreatedAt = message.CreatedAt ?? DateTimeOffset.UtcNow,
            MessageId = message.MessageId ?? Guid.NewGuid().ToString("N"),
            ResponseId = responseId,
            RawRepresentation = raw
        };
-
-        this.ChatHistoryProvider.AddMessages(this, update.ToChatMessage());
-
-        return update;
    }

-    private async ValueTask<StreamingRun> CreateOrResumeRunAsync(List<ChatMessage> messages, CancellationToken cancellationToken = default)
+    private async ValueTask<ResumeRunResult> CreateOrResumeRunAsync(List<ChatMessage> messages, CancellationToken cancellationToken = default)
    {
        // The workflow is validated to be a ChatProtocol workflow by the WorkflowHostAgent before creating the session,
        // and does not need to be checked again here.
@@ -154,109 +167,257 @@ internal sealed class WorkflowSession : AgentSession
                                               cancellationToken)
                            .ConfigureAwait(false);

-            await run.TrySendMessageAsync(messages).ConfigureAwait(false);
-            return run;
+            // Process messages: convert response content to ExternalResponse, send regular messages as-is
+            ResumeDispatchInfo dispatchInfo = await this.SendMessagesWithResponseConversionAsync(run, messages).ConfigureAwait(false);
+            return new ResumeRunResult(run, dispatchInfo);
        }

-        return await this._executionEnvironment
+        StreamingRun newRun = await this._executionEnvironment
                            .RunStreamingAsync(this._workflow,
                                         messages,
                                         this.SessionId,
                                         cancellationToken)
                            .ConfigureAwait(false);
+        return new ResumeRunResult(newRun);
    }

+    /// <summary>
+    /// Sends messages to the run, converting FunctionResultContent and UserInputResponseContent
+    /// to ExternalResponse when there's a matching pending request.
+    /// </summary>
+    /// <returns>
+    /// Structured information about how resume content was dispatched.
+    /// </returns>
+    private async ValueTask<ResumeDispatchInfo> SendMessagesWithResponseConversionAsync(StreamingRun run, List<ChatMessage> messages)
+    {
+        List<ChatMessage> regularMessages = [];
+        // Responses are deferred until after regular messages are queued so response handlers
+        // can merge buffered regular content in the same continuation turn.
+        List<(ExternalResponse Response, string RequestId)> externalResponses = [];
+        bool hasMatchedResponseForStartExecutor = false;
+
+        // Tracks content IDs already matched to pending requests within this invocation,
+        // preventing duplicate responses for the same ID from being sent to the workflow engine.
+        HashSet<string>? matchedContentIds = null;
+
+        foreach (ChatMessage message in messages)
+        {
+            List<AIContent> regularContents = [];
+
+            foreach (AIContent content in message.Contents)
+            {
+                string? contentId = GetResponseContentId(content);
+
+                // Skip duplicate response content for an already-matched content ID
+                if (contentId != null && matchedContentIds?.Contains(contentId) == true)
+                {
+                    continue;
+                }
+
+                if (contentId != null
+                    && this.TryGetPendingRequest(contentId) is ExternalRequest pendingRequest)
+                {
+                    // For intercepted/complex topologies the port may not be registered in the EdgeMap.
+                    // Treat unknown port as non-start-executor (conservative): TurnToken will still be sent.
+                    if (run.TryGetResponsePortExecutorId(pendingRequest.PortInfo.PortId, out string? responseExecutorId))
+                    {
+                        hasMatchedResponseForStartExecutor |= string.Equals(responseExecutorId, this._workflow.StartExecutorId, StringComparison.Ordinal);
+                    }
+
+                    AIContent normalizedResponseContent = NormalizeResponseContentForDelivery(content, pendingRequest);
+                    externalResponses.Add((pendingRequest.CreateResponse(normalizedResponseContent), pendingRequest.RequestId));
+                    (matchedContentIds ??= new(StringComparer.Ordinal)).Add(contentId);
+                }
+                else
+                {
+                    regularContents.Add(content);
+                }
+            }
+
+            if (regularContents.Count > 0)
+            {
+                ChatMessage cloned = message.Clone();
+                cloned.Contents = regularContents;
+                regularMessages.Add(cloned);
+            }
+        }
+
+        // Send regular messages first so response handlers can merge them with responses.
+        bool hasRegularMessages = regularMessages.Count > 0;
+        if (hasRegularMessages)
+        {
+            await run.TrySendMessageAsync(regularMessages).ConfigureAwait(false);
+        }
+
+        // Send external responses after regular messages.
+        bool hasMatchedExternalResponses = false;
+        foreach ((ExternalResponse response, string requestId) in externalResponses)
+        {
+            await run.SendResponseAsync(response).ConfigureAwait(false);
+            hasMatchedExternalResponses = true;
+            this.RemovePendingRequest(requestId);
+        }
+
+        return new ResumeDispatchInfo(
+            hasRegularMessages,
+            hasMatchedExternalResponses,
+            hasMatchedResponseForStartExecutor);
+    }
+
+    /// <summary>
+    /// Creates the workflow-facing request content surfaced in response updates.
+    /// </summary>
+    private static AIContent CreateRequestContentForDelivery(ExternalRequest request) => request switch
+    {
+        ExternalRequest externalRequest when externalRequest.TryGetDataAs(out FunctionCallContent? functionCallContent)
+            => CloneFunctionCallContent(functionCallContent, externalRequest.RequestId),
+        ExternalRequest externalRequest when externalRequest.TryGetDataAs(out ToolApprovalRequestContent? toolApprovalRequestContent)
+            => CloneToolApprovalRequestContent(toolApprovalRequestContent, externalRequest.RequestId),
+        ExternalRequest externalRequest
+            => externalRequest.ToFunctionCall(),
+    };
+
+    /// <summary>
+    /// Rewrites workflow-facing response content back to the original agent-owned content ID.
+    /// </summary>
+    private static AIContent NormalizeResponseContentForDelivery(AIContent content, ExternalRequest request) => content switch
+    {
+        FunctionResultContent functionResultContent when request.TryGetDataAs(out FunctionCallContent? functionCallContent)
+            => CloneFunctionResultContent(functionResultContent, functionCallContent.CallId),
+        ToolApprovalResponseContent toolApprovalResponseContent when request.TryGetDataAs(out ToolApprovalRequestContent? toolApprovalRequestContent)
+            => CloneToolApprovalResponseContent(toolApprovalResponseContent, toolApprovalRequestContent.RequestId),
+        _ => content,
+    };
+
+    /// <summary>
+    /// Gets the workflow-facing request ID from response content types.
+    /// </summary>
+    private static string? GetResponseContentId(AIContent content) => content switch
+    {
+        FunctionResultContent functionResultContent => functionResultContent.CallId,
+        ToolApprovalResponseContent toolApprovalResponseContent => toolApprovalResponseContent.RequestId,
+        _ => null
+    };
+
+    /// <summary>
+    /// Tries to get a pending request by workflow-facing request ID.
+    /// </summary>
+    private ExternalRequest? TryGetPendingRequest(string requestId) =>
+        this._pendingRequests.TryGetValue(requestId, out ExternalRequest? request) ? request : null;
+
+    /// <summary>
+    /// Adds a pending request indexed by workflow-facing request ID.
+    /// </summary>
+    private void AddPendingRequest(string requestId, ExternalRequest request) => this._pendingRequests[requestId] = request;
+
+    /// <summary>
+    /// Removes a pending request by workflow-facing request ID.
+    /// </summary>
+    private void RemovePendingRequest(string requestId) =>
+        this._pendingRequests.Remove(requestId);
+
    internal async
    IAsyncEnumerable<AgentResponseUpdate> InvokeStageAsync(
        [EnumeratorCancellation] CancellationToken cancellationToken = default)
    {
-        try
-        {
-            this.LastResponseId = Guid.NewGuid().ToString("N");
-            List<ChatMessage> messages = this.ChatHistoryProvider.GetFromBookmark(this).ToList();
+        this.LastResponseId = Guid.NewGuid().ToString("N");
+        List<ChatMessage> messages = this.ChatHistoryProvider.GetFromBookmark(this).ToList();

-#pragma warning disable CA2007 // Analyzer misfiring and not seeing .ConfigureAwait(false) below.
-            await using StreamingRun run =
-                await this.CreateOrResumeRunAsync(messages, cancellationToken).ConfigureAwait(false);
+        ResumeRunResult resumeResult =
+            await this.CreateOrResumeRunAsync(messages, cancellationToken).ConfigureAwait(false);
+
+#pragma warning disable CA2007 // Analyzer misfiring.
+        await using StreamingRun run = resumeResult.Run;
 #pragma warning restore CA2007

+        ResumeDispatchInfo dispatchInfo = resumeResult.DispatchInfo;
+
+        // Send a TurnToken to the start executor unless the only activity is an external
+        // response directed at the start executor itself (which self-emits a TurnToken via
+        // ContinueTurnAsync). Non-start executors (e.g., RequestInfoExecutor) do not emit
+        // TurnTokens after processing responses, so the session must always provide one.
+        bool shouldSendTurnToken =
+            !dispatchInfo.HasMatchedExternalResponses
+            || !dispatchInfo.HasMatchedResponseForStartExecutor;
+        if (shouldSendTurnToken)
+        {
            await run.TrySendMessageAsync(new TurnToken(emitEvents: true)).ConfigureAwait(false);
-            await foreach (WorkflowEvent evt in run.WatchStreamAsync(blockOnPendingRequest: false, cancellationToken)
+        }
+        await foreach (WorkflowEvent evt in run.WatchStreamAsync(blockOnPendingRequest: false, cancellationToken)
                                               .ConfigureAwait(false)
                                               .WithCancellation(cancellationToken))
-            {
-                switch (evt)
-                {
-                    case AgentResponseUpdateEvent agentUpdate:
-                        yield return agentUpdate.Update;
-                        break;
-
-                    case RequestInfoEvent requestInfo:
-                        FunctionCallContent fcContent = requestInfo.Request.ToFunctionCall();
-                        AgentResponseUpdate update = this.CreateUpdate(this.LastResponseId, evt, fcContent);
-                        yield return update;
-                        break;
-
-                    case WorkflowErrorEvent workflowError:
-                        Exception? exception = workflowError.Exception;
-                        if (exception is TargetInvocationException tie && tie.InnerException != null)
-                        {
-                            exception = tie.InnerException;
-                        }
-
-                        if (exception != null)
-                        {
-                            string message = this._includeExceptionDetails
-                                           ? exception.Message
-                                           : "An error occurred while executing the workflow.";
-
-                            ErrorContent errorContent = new(message);
-                            yield return this.CreateUpdate(this.LastResponseId, evt, errorContent);
-                        }
-
-                        break;
-
-                    case SuperStepCompletedEvent stepCompleted:
-                        this.LastCheckpoint = stepCompleted.CompletionInfo?.Checkpoint;
-                        goto default;
-
-                    case WorkflowOutputEvent output:
-                        IEnumerable<ChatMessage>? updateMessages = output.Data switch
-                        {
-                            IEnumerable<ChatMessage> chatMessages => chatMessages,
-                            ChatMessage chatMessage => [chatMessage],
-                            _ => null
-                        };
-
-                        if (!this._includeWorkflowOutputsInResponse || updateMessages == null)
-                        {
-                            goto default;
-                        }
-
-                        foreach (ChatMessage message in updateMessages)
-                        {
-                            yield return this.CreateUpdate(this.LastResponseId, evt, message);
-                        }
-                        break;
-
-                    default:
-                        // Emit all other workflow events for observability (DevUI, logging, etc.)
-                        yield return new AgentResponseUpdate(ChatRole.Assistant, [])
-                        {
-                            CreatedAt = DateTimeOffset.UtcNow,
-                            MessageId = Guid.NewGuid().ToString("N"),
-                            Role = ChatRole.Assistant,
-                            ResponseId = this.LastResponseId,
-                            RawRepresentation = evt
-                        };
-                        break;
-                }
-            }
-        }
-        finally
        {
-            // Do we want to try to undo the step, and not update the bookmark?
-            this.ChatHistoryProvider.UpdateBookmark(this);
+            switch (evt)
+            {
+                case AgentResponseUpdateEvent agentUpdate:
+                    yield return agentUpdate.Update;
+                    break;
+
+                case RequestInfoEvent requestInfo:
+                    AIContent requestContent = CreateRequestContentForDelivery(requestInfo.Request);
+
+                    // Track the pending request so we can convert incoming responses back to ExternalResponse.
+                    // External callers respond using the workflow-facing request ID, which is always RequestId.
+                    this.AddPendingRequest(requestInfo.Request.RequestId, requestInfo.Request);
+
+                    AgentResponseUpdate update = this.CreateUpdate(this.LastResponseId, evt, requestContent);
+                    yield return update;
+                    break;
+
+                case WorkflowErrorEvent workflowError:
+                    Exception? exception = workflowError.Exception;
+                    if (exception is TargetInvocationException tie && tie.InnerException != null)
+                    {
+                        exception = tie.InnerException;
+                    }
+
+                    if (exception != null)
+                    {
+                        string message = this._includeExceptionDetails
+                                       ? exception.Message
+                                       : "An error occurred while executing the workflow.";
+
+                        ErrorContent errorContent = new(message);
+                        yield return this.CreateUpdate(this.LastResponseId, evt, errorContent);
+                    }
+
+                    break;
+
+                case SuperStepCompletedEvent stepCompleted:
+                    this.LastCheckpoint = stepCompleted.CompletionInfo?.Checkpoint;
+                    goto default;
+
+                case WorkflowOutputEvent output:
+                    IEnumerable<ChatMessage>? updateMessages = output.Data switch
+                    {
+                        IEnumerable<ChatMessage> chatMessages => chatMessages,
+                        ChatMessage chatMessage => [chatMessage],
+                        _ => null
+                    };
+
+                    if (!this._includeWorkflowOutputsInResponse || updateMessages == null)
+                    {
+                        goto default;
+                    }
+
+                    foreach (ChatMessage message in updateMessages)
+                    {
+                        yield return this.CreateUpdate(this.LastResponseId, evt, message);
+                    }
+                    break;
+
+                default:
+                    // Emit all other workflow events for observability (DevUI, logging, etc.)
+                    yield return new AgentResponseUpdate(ChatRole.Assistant, [])
+                    {
+                        CreatedAt = DateTimeOffset.UtcNow,
+                        MessageId = Guid.NewGuid().ToString("N"),
+                        Role = ChatRole.Assistant,
+                        ResponseId = this.LastResponseId,
+                        RawRepresentation = evt
+                    };
+                    break;
+            }
        }
    }

@@ -267,15 +428,116 @@ internal sealed class WorkflowSession : AgentSession
    /// <inheritdoc/>
    public WorkflowChatHistoryProvider ChatHistoryProvider { get; }

+    /// <summary>
+    /// Captures the outcome of creating or resuming a workflow run,
+    /// indicating what types of messages were sent during resume.
+    /// </summary>
+    private readonly struct ResumeRunResult
+    {
+        /// <summary>The streaming run that was created or resumed.</summary>
+        public StreamingRun Run { get; }
+
+        /// <summary>How resume-time content was dispatched into the workflow runtime.</summary>
+        public ResumeDispatchInfo DispatchInfo { get; }
+
+        public ResumeRunResult(StreamingRun run, ResumeDispatchInfo dispatchInfo = default)
+        {
+            this.Run = Throw.IfNull(run);
+            this.DispatchInfo = dispatchInfo;
+        }
+    }
+
+    /// <summary>
+    /// Captures how resumed input was split across regular-message and external-response delivery paths.
+    /// </summary>
+    private readonly struct ResumeDispatchInfo
+    {
+        public ResumeDispatchInfo(bool hasRegularMessages, bool hasMatchedExternalResponses, bool hasMatchedResponseForStartExecutor)
+        {
+            this.HasRegularMessages = hasRegularMessages;
+            this.HasMatchedExternalResponses = hasMatchedExternalResponses;
+            this.HasMatchedResponseForStartExecutor = hasMatchedResponseForStartExecutor;
+        }
+
+        public bool HasRegularMessages { get; }
+
+        public bool HasMatchedExternalResponses { get; }
+
+        public bool HasMatchedResponseForStartExecutor { get; }
+    }
+
+    /// <summary>
+    /// Clones a <see cref="FunctionCallContent"/> with a workflow-facing call ID.
+    /// </summary>
+    private static FunctionCallContent CloneFunctionCallContent(FunctionCallContent content, string callId)
+    {
+        FunctionCallContent clone = new(callId, content.Name, content.Arguments)
+        {
+            Exception = content.Exception,
+            InformationalOnly = content.InformationalOnly,
+        };
+
+        return CopyContentMetadata(content, clone);
+    }
+
+    /// <summary>
+    /// Clones a <see cref="FunctionResultContent"/> with an agent-owned call ID.
+    /// </summary>
+    private static FunctionResultContent CloneFunctionResultContent(FunctionResultContent content, string callId)
+    {
+        FunctionResultContent clone = new(callId, content.Result)
+        {
+            Exception = content.Exception,
+        };
+
+        return CopyContentMetadata(content, clone);
+    }
+
+    /// <summary>
+    /// Clones a <see cref="ToolApprovalRequestContent"/> with a workflow-facing request ID.
+    /// </summary>
+    private static ToolApprovalRequestContent CloneToolApprovalRequestContent(ToolApprovalRequestContent content, string id)
+    {
+        ToolApprovalRequestContent clone = new(id, content.ToolCall);
+        return CopyContentMetadata(content, clone);
+    }
+
+    /// <summary>
+    /// Clones a <see cref="ToolApprovalResponseContent"/> with an agent-owned request ID.
+    /// </summary>
+    private static ToolApprovalResponseContent CloneToolApprovalResponseContent(ToolApprovalResponseContent content, string id)
+    {
+        ToolApprovalResponseContent clone = new(id, content.Approved, content.ToolCall)
+        {
+            Reason = content.Reason,
+        };
+
+        return CopyContentMetadata(content, clone);
+    }
+
+    /// <summary>
+    /// Copies shared <see cref="AIContent"/> metadata to a cloned content instance.
+    /// </summary>
+    private static TContent CopyContentMetadata<TContent>(AIContent source, TContent target)
+        where TContent : AIContent
+    {
+        target.AdditionalProperties = source.AdditionalProperties;
+        target.Annotations = source.Annotations;
+        target.RawRepresentation = source.RawRepresentation;
+        return target;
+    }
+
    internal sealed class SessionState(
        string sessionId,
        CheckpointInfo? lastCheckpoint,
        InMemoryCheckpointManager? checkpointManager = null,
-        AgentSessionStateBag? stateBag = null)
+        AgentSessionStateBag? stateBag = null,
+        Dictionary<string, ExternalRequest>? pendingRequests = null)
    {
        public string SessionId { get; } = sessionId;
        public CheckpointInfo? LastCheckpoint { get; } = lastCheckpoint;
        public InMemoryCheckpointManager? CheckpointManager { get; } = checkpointManager;
        public AgentSessionStateBag StateBag { get; } = stateBag ?? new();
+        public Dictionary<string, ExternalRequest>? PendingRequests { get; } = pendingRequests;
    }
 }
@@ -71,6 +71,7 @@ internal static partial class WorkflowsJsonUtilities
    [JsonSerializable(typeof(PortableValue))]
    [JsonSerializable(typeof(PortableMessageEnvelope))]
    [JsonSerializable(typeof(InMemoryCheckpointManager))]
+    [JsonSerializable(typeof(CheckpointFileIndexEntry))]

    // Runtime State Types
    [JsonSerializable(typeof(ScopeKey))]
@@ -161,6 +161,8 @@ internal sealed class AIContextProviderChatClient : DelegatingChatClient
        }

        // Materialize the accumulated context back into messages and options.
+        // Clone options to avoid mutating the caller's instance across calls.
+        options = options?.Clone();
        var enrichedMessages = aiContext.Messages ?? [];

        var tools = aiContext.Tools as IList<AITool> ?? aiContext.Tools?.ToList();
@@ -138,6 +138,9 @@ public sealed partial class ChatClientAgent : AIAgent
        this._aiContextProviderStateKeys = ValidateAndCollectStateKeys(this._agentOptions?.AIContextProviders, this.ChatHistoryProvider);

        this._logger = (loggerFactory ?? chatClient.GetService<ILoggerFactory>() ?? NullLoggerFactory.Instance).CreateLogger<ChatClientAgent>();
+
+        // Warn if using a custom chat client stack with end-of-run persistence but no ChatHistoryPersistingChatClient.
+        this.WarnOnMissingPersistingClient();
    }

    /// <summary>
@@ -211,12 +214,14 @@ public sealed partial class ChatClientAgent : AIAgent
         ChatClientAgentContinuationToken? _) =
            await this.PrepareSessionAndMessagesAsync(session, inputMessages, options, cancellationToken).ConfigureAwait(false);

-        var chatClient = this.ChatClient;
+        // Update the run context with the resolved session so any downstream classes
+        // always have a valid session, even when the caller passed null.
+        EnsureRunContextHasSession(safeSession);

+        var chatClient = this.ChatClient;
        chatClient = ApplyRunOptionsTransformations(options, chatClient);

        var loggingAgentName = this.GetLoggingAgentName();
-
        this._logger.LogAgentChatClientInvokingAgent(nameof(RunAsync), this.Id, loggingAgentName, this._chatClientType);

        // Call the IChatClient and notify the AIContextProvider of any failures.
@@ -227,8 +232,7 @@ public sealed partial class ChatClientAgent : AIAgent
        }
        catch (Exception ex)
        {
-            await this.NotifyChatHistoryProviderOfFailureAsync(safeSession, ex, inputMessagesForChatClient, chatOptions, cancellationToken).ConfigureAwait(false);
-            await this.NotifyAIContextProviderOfFailureAsync(safeSession, ex, inputMessagesForChatClient, cancellationToken).ConfigureAwait(false);
+            await this.NotifyProvidersOfFailureAtEndOfRunAsync(safeSession, ex, inputMessagesForChatClient, chatOptions, cancellationToken).ConfigureAwait(false);
            throw;
        }

@@ -236,7 +240,8 @@ public sealed partial class ChatClientAgent : AIAgent

        // We can derive the type of supported session from whether we have a conversation id,
        // so let's update it and set the conversation id for the service session case.
-        this.UpdateSessionConversationId(safeSession, chatResponse.ConversationId, cancellationToken);
+        var forceEndOfRunPersistence = chatOptions?.ContinuationToken is not null || chatOptions?.AllowBackgroundResponses is true;
+        this.UpdateSessionConversationIdAtEndOfRun(safeSession, chatResponse.ConversationId, cancellationToken, forceUpdate: forceEndOfRunPersistence);

        // Ensure that the author name is set for each message in the response.
        foreach (ChatMessage chatResponseMessage in chatResponse.Messages)
@@ -244,11 +249,10 @@ public sealed partial class ChatClientAgent : AIAgent
            chatResponseMessage.AuthorName ??= this.Name;
        }

-        // Only notify the session of new messages if the chatResponse was successful to avoid inconsistent message state in the session.
-        await this.NotifyChatHistoryProviderOfNewMessagesAsync(safeSession, inputMessagesForChatClient, chatResponse.Messages, chatOptions, cancellationToken).ConfigureAwait(false);
-
-        // Notify the AIContextProvider of all new messages.
-        await this.NotifyAIContextProviderOfSuccessAsync(safeSession, inputMessagesForChatClient, chatResponse.Messages, cancellationToken).ConfigureAwait(false);
+        // Notify providers of all new messages unless persistence is handled per-service-call by the decorator.
+        // When background responses are allowed, force notification since per-service-call persistence
+        // is unreliable (the caller may stop consuming the stream before the decorator can persist).
+        await this.NotifyProvidersOfNewMessagesAtEndOfRunAsync(safeSession, inputMessagesForChatClient, chatResponse.Messages, chatOptions, cancellationToken, forceNotify: forceEndOfRunPersistence).ConfigureAwait(false);

        return new AgentResponse(chatResponse)
        {
@@ -296,6 +300,10 @@ public sealed partial class ChatClientAgent : AIAgent
         ChatClientAgentContinuationToken? continuationToken) =
            await this.PrepareSessionAndMessagesAsync(session, inputMessages, options, cancellationToken).ConfigureAwait(false);

+        // Update the run context with the resolved session so any downstream classes
+        // always have a valid session, even when the caller passed null.
+        EnsureRunContextHasSession(safeSession);
+
        var chatClient = this.ChatClient;

        chatClient = ApplyRunOptionsTransformations(options, chatClient);
@@ -315,8 +323,7 @@ public sealed partial class ChatClientAgent : AIAgent
        }
        catch (Exception ex)
        {
-            await this.NotifyChatHistoryProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
-            await this.NotifyAIContextProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), cancellationToken).ConfigureAwait(false);
+            await this.NotifyProvidersOfFailureAtEndOfRunAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
            throw;
        }

@@ -330,8 +337,7 @@ public sealed partial class ChatClientAgent : AIAgent
        }
        catch (Exception ex)
        {
-            await this.NotifyChatHistoryProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
-            await this.NotifyAIContextProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), cancellationToken).ConfigureAwait(false);
+            await this.NotifyProvidersOfFailureAtEndOfRunAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
            throw;
        }

@@ -353,27 +359,31 @@ public sealed partial class ChatClientAgent : AIAgent

            try
            {
+                // Re-ensure the run context has the resolved session before each MoveNextAsync.
+                // The base class RunStreamingAsync restores the original context (potentially with
+                // null session) after each yield, so we must re-establish it for the decorator.
+                EnsureRunContextHasSession(safeSession);
                hasUpdates = await responseUpdatesEnumerator.MoveNextAsync().ConfigureAwait(false);
            }
            catch (Exception ex)
            {
-                await this.NotifyChatHistoryProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
-                await this.NotifyAIContextProviderOfFailureAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), cancellationToken).ConfigureAwait(false);
+                await this.NotifyProvidersOfFailureAtEndOfRunAsync(safeSession, ex, GetInputMessages(inputMessagesForChatClient, continuationToken), chatOptions, cancellationToken).ConfigureAwait(false);
                throw;
            }
        }

        var chatResponse = responseUpdates.ToChatResponse();

+        var forceEndOfRunPersistence = continuationToken is not null || chatOptions?.AllowBackgroundResponses is true;
+
        // We can derive the type of supported session from whether we have a conversation id,
        // so let's update it and set the conversation id for the service session case.
-        this.UpdateSessionConversationId(safeSession, chatResponse.ConversationId, cancellationToken);
+        this.UpdateSessionConversationIdAtEndOfRun(safeSession, chatResponse.ConversationId, cancellationToken, forceUpdate: forceEndOfRunPersistence);

-        // To avoid inconsistent state we only notify the session of the input messages if no error occurs after the initial request.
-        await this.NotifyChatHistoryProviderOfNewMessagesAsync(safeSession, GetInputMessages(inputMessagesForChatClient, continuationToken), chatResponse.Messages, chatOptions, cancellationToken).ConfigureAwait(false);
-
-        // Notify the AIContextProvider of all new messages.
-        await this.NotifyAIContextProviderOfSuccessAsync(safeSession, GetInputMessages(inputMessagesForChatClient, continuationToken), chatResponse.Messages, cancellationToken).ConfigureAwait(false);
+        // Notify providers of all new messages unless persistence is handled per-service-call by the decorator.
+        // When resuming from a continuation token or using background responses, force notification
+        // to send the combined data (per-service-call persistence is unreliable for these scenarios).
+        await this.NotifyProvidersOfNewMessagesAtEndOfRunAsync(safeSession, GetInputMessages(inputMessagesForChatClient, continuationToken), chatResponse.Messages, chatOptions, cancellationToken, forceNotify: forceEndOfRunPersistence).ConfigureAwait(false);
    }

    /// <inheritdoc/>
@@ -441,17 +451,29 @@ public sealed partial class ChatClientAgent : AIAgent
    #region Private

    /// <summary>
-    /// Notify the <see cref="AIContextProvider"/> when an agent run succeeded, if there is an <see cref="AIContextProvider"/>.
+    /// Notifies the <see cref="ChatHistoryProvider"/> and all <see cref="AIContextProviders"/> of successfully completed messages.
    /// </summary>
-    private async Task NotifyAIContextProviderOfSuccessAsync(
+    /// <remarks>
+    /// This method is also called by <see cref="ChatHistoryPersistingChatClient"/> to persist messages per-service-call.
+    /// </remarks>
+    internal async Task NotifyProvidersOfNewMessagesAsync(
        ChatClientAgentSession session,
-        IEnumerable<ChatMessage> inputMessages,
+        IEnumerable<ChatMessage> requestMessages,
        IEnumerable<ChatMessage> responseMessages,
+        ChatOptions? chatOptions,
        CancellationToken cancellationToken)
    {
+        ChatHistoryProvider? chatHistoryProvider = this.ResolveChatHistoryProvider(chatOptions, session);
+
+        if (chatHistoryProvider is not null)
+        {
+            var invokedContext = new ChatHistoryProvider.InvokedContext(this, session, requestMessages, responseMessages);
+            await chatHistoryProvider.InvokedAsync(invokedContext, cancellationToken).ConfigureAwait(false);
+        }
+
        if (this.AIContextProviders is { Count: > 0 } contextProviders)
        {
-            AIContextProvider.InvokedContext invokedContext = new(this, session, inputMessages, responseMessages);
+            AIContextProvider.InvokedContext invokedContext = new(this, session, requestMessages, responseMessages);

            foreach (var contextProvider in contextProviders)
            {
@@ -461,17 +483,29 @@ public sealed partial class ChatClientAgent : AIAgent
    }

    /// <summary>
-    /// Notify the <see cref="AIContextProvider"/> of any failure during an agent run, if there is an <see cref="AIContextProvider"/>.
+    /// Notifies the <see cref="ChatHistoryProvider"/> and all <see cref="AIContextProviders"/> of a failure during a service call.
    /// </summary>
-    private async Task NotifyAIContextProviderOfFailureAsync(
+    /// <remarks>
+    /// This method is also called by <see cref="ChatHistoryPersistingChatClient"/> to report failures per-service-call.
+    /// </remarks>
+    internal async Task NotifyProvidersOfFailureAsync(
        ChatClientAgentSession session,
        Exception ex,
-        IEnumerable<ChatMessage> inputMessages,
+        IEnumerable<ChatMessage> requestMessages,
+        ChatOptions? chatOptions,
        CancellationToken cancellationToken)
    {
+        ChatHistoryProvider? chatHistoryProvider = this.ResolveChatHistoryProvider(chatOptions, session);
+
+        if (chatHistoryProvider is not null)
+        {
+            var invokedContext = new ChatHistoryProvider.InvokedContext(this, session, requestMessages, ex);
+            await chatHistoryProvider.InvokedAsync(invokedContext, cancellationToken).ConfigureAwait(false);
+        }
+
        if (this.AIContextProviders is { Count: > 0 } contextProviders)
        {
-            AIContextProvider.InvokedContext invokedContext = new(this, session, inputMessages, ex);
+            AIContextProvider.InvokedContext invokedContext = new(this, session, requestMessages, ex);

            foreach (var contextProvider in contextProviders)
            {
@@ -667,6 +701,12 @@ public sealed partial class ChatClientAgent : AIAgent
            throw new InvalidOperationException("A session must be provided when continuing a background response with a continuation token.");
        }

+        if ((continuationToken is not null || chatOptions?.AllowBackgroundResponses is true) && this.PersistsChatHistoryPerServiceCall && this._logger.IsEnabled(LogLevel.Warning))
+        {
+            var warningAgentName = this.GetLoggingAgentName();
+            this._logger.LogAgentChatClientBackgroundResponseFallback(this.Id, warningAgentName);
+        }
+
        session ??= await this.CreateSessionAsync(cancellationToken).ConfigureAwait(false);
        if (session is not ChatClientAgentSession typedSession)
        {
@@ -754,7 +794,7 @@ public sealed partial class ChatClientAgent : AIAgent
        return (typedSession, chatOptions, messagesList, continuationToken);
    }

-    private void UpdateSessionConversationId(ChatClientAgentSession session, string? responseConversationId, CancellationToken cancellationToken)
+    internal void UpdateSessionConversationId(ChatClientAgentSession session, string? responseConversationId, CancellationToken cancellationToken)
    {
        if (string.IsNullOrWhiteSpace(responseConversationId) && !string.IsNullOrWhiteSpace(session.ConversationId))
        {
@@ -798,45 +838,162 @@ public sealed partial class ChatClientAgent : AIAgent
        }
    }

-    private Task NotifyChatHistoryProviderOfFailureAsync(
+    /// <summary>
+    /// Updates the session conversation ID at the end of an agent run.
+    /// </summary>
+    /// <remarks>
+    /// When a <see cref="ChatHistoryPersistingChatClient"/> in persist mode handles per-service-call
+    /// conversation ID updates, this end-of-run update is skipped. When the decorator is in mark-only
+    /// mode or absent, the update is performed here. When <paramref name="forceUpdate"/> is <see langword="true"/>
+    /// (continuation token scenarios), the update is always performed.
+    /// </remarks>
+    private void UpdateSessionConversationIdAtEndOfRun(ChatClientAgentSession session, string? responseConversationId, CancellationToken cancellationToken, bool forceUpdate = false)
+    {
+        if (!forceUpdate && this.PersistsChatHistoryPerServiceCall)
+        {
+            return;
+        }
+
+        this.UpdateSessionConversationId(session, responseConversationId, cancellationToken);
+    }
+
+    /// <summary>
+    /// Notifies providers of successfully completed messages at the end of an agent run.
+    /// </summary>
+    /// <remarks>
+    /// When a <see cref="ChatHistoryPersistingChatClient"/> in persist mode handles per-service-call
+    /// notification, this end-of-run notification is skipped. When the decorator is in mark-only mode,
+    /// only the marked messages are persisted. When no decorator is present (custom stack with
+    /// <see cref="ChatClientAgentOptions.PersistChatHistoryAtEndOfRun"/>), all messages are persisted.
+    /// When <paramref name="forceNotify"/> is <see langword="true"/> (continuation token or
+    /// background response scenarios), notification is always performed with all messages because
+    /// per-service-call persistence is unreliable in these scenarios.
+    /// </remarks>
+    private Task NotifyProvidersOfNewMessagesAtEndOfRunAsync(
+        ChatClientAgentSession session,
+        IEnumerable<ChatMessage> requestMessages,
+        IEnumerable<ChatMessage> responseMessages,
+        ChatOptions? chatOptions,
+        CancellationToken cancellationToken,
+        bool forceNotify = false)
+    {
+        if (!forceNotify && this.PersistsChatHistoryPerServiceCall)
+        {
+            return Task.CompletedTask;
+        }
+
+        if (!forceNotify && this.HasMarkOnlyChatHistoryPersistingClient)
+        {
+            // In mark-only mode, persist only messages that were marked by the decorator.
+            var markedRequestMessages = GetMarkedMessages(requestMessages);
+            var markedResponseMessages = GetMarkedMessages(responseMessages);
+            return this.NotifyProvidersOfNewMessagesAsync(session, markedRequestMessages, markedResponseMessages, chatOptions, cancellationToken);
+        }
+
+        return this.NotifyProvidersOfNewMessagesAsync(session, requestMessages, responseMessages, chatOptions, cancellationToken);
+    }
+
+    /// <summary>
+    /// Notifies providers of a failure at the end of an agent run.
+    /// </summary>
+    /// <remarks>
+    /// When a <see cref="ChatHistoryPersistingChatClient"/> in persist mode handles per-service-call
+    /// notification (including failure), this end-of-run notification is skipped to avoid
+    /// duplicate notification. In all other cases, failure is reported at the end of the run.
+    /// </remarks>
+    private Task NotifyProvidersOfFailureAtEndOfRunAsync(
        ChatClientAgentSession session,
        Exception ex,
        IEnumerable<ChatMessage> requestMessages,
        ChatOptions? chatOptions,
        CancellationToken cancellationToken)
    {
-        ChatHistoryProvider? provider = this.ResolveChatHistoryProvider(chatOptions, session);
-
-        // Only notify the provider if we have one.
-        // If we don't have one, it means that the chat history is service managed and the underlying service is responsible for storing messages.
-        if (provider is not null)
+        if (this.PersistsChatHistoryPerServiceCall)
        {
-            var invokedContext = new ChatHistoryProvider.InvokedContext(this, session, requestMessages, ex);
-
-            return provider.InvokedAsync(invokedContext, cancellationToken).AsTask();
+            return Task.CompletedTask;
        }

-        return Task.CompletedTask;
+        return this.NotifyProvidersOfFailureAsync(session, ex, requestMessages, chatOptions, cancellationToken);
    }

-    private Task NotifyChatHistoryProviderOfNewMessagesAsync(
-        ChatClientAgentSession session,
-        IEnumerable<ChatMessage> requestMessages,
-        IEnumerable<ChatMessage> responseMessages,
-        ChatOptions? chatOptions,
-        CancellationToken cancellationToken)
+    /// <summary>
+    /// Gets a value indicating whether the agent has a <see cref="ChatHistoryPersistingChatClient"/>
+    /// decorator in persist mode (not mark-only), which handles per-service-call persistence.
+    /// </summary>
+    private bool PersistsChatHistoryPerServiceCall
    {
-        ChatHistoryProvider? provider = this.ResolveChatHistoryProvider(chatOptions, session);
-
-        // Only notify the provider if we have one.
-        // If we don't have one, it means that the chat history is service managed and the underlying service is responsible for storing messages.
-        if (provider is not null)
+        get
        {
-            var invokedContext = new ChatHistoryProvider.InvokedContext(this, session, requestMessages, responseMessages);
-            return provider.InvokedAsync(invokedContext, cancellationToken).AsTask();
+            var persistingClient = this.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+            return persistingClient?.MarkOnly == false;
+        }
+    }
+
+    /// <summary>
+    /// Gets a value indicating whether the agent has a <see cref="ChatHistoryPersistingChatClient"/>
+    /// decorator in mark-only mode, which marks messages for later persistence at the end of the run.
+    /// </summary>
+    private bool HasMarkOnlyChatHistoryPersistingClient
+    {
+        get
+        {
+            var persistingClient = this.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+            return persistingClient?.MarkOnly == true;
+        }
+    }
+
+    /// <summary>
+    /// Returns only the messages that have been marked as persisted by a <see cref="ChatHistoryPersistingChatClient"/> in mark-only mode.
+    /// </summary>
+    private static List<ChatMessage> GetMarkedMessages(IEnumerable<ChatMessage> messages)
+    {
+        return messages.Where(m =>
+            m.AdditionalProperties?.TryGetValue(ChatHistoryPersistingChatClient.PersistedMarkerKey, out var value) == true && value is true).ToList();
+    }
+
+    /// <summary>
+    /// Ensures that <see cref="AIAgent.CurrentRunContext"/> contains the resolved session.
+    /// </summary>
+    /// <remarks>
+    /// The base class sets <see cref="AIAgent.CurrentRunContext"/> with the raw session parameter
+    /// (which may be null) and restores it after each yield in streaming scenarios. After
+    /// <see cref="PrepareSessionAndMessagesAsync"/> resolves or creates a session, we update the
+    /// context so the <see cref="ChatHistoryPersistingChatClient"/> decorator always has a valid session.
+    /// The original agent from the context is preserved to maintain the top-of-stack agent in
+    /// decorated agent scenarios.
+    /// </remarks>
+    private static void EnsureRunContextHasSession(ChatClientAgentSession safeSession)
+    {
+        var context = CurrentRunContext;
+        if (context is not null && context.Session != safeSession)
+        {
+            CurrentRunContext = new(context.Agent, safeSession, context.RequestMessages, context.RunOptions);
+        }
+    }
+
+    /// <summary>
+    /// Checks for potential misconfiguration when using a custom chat client stack and logs warnings.
+    /// </summary>
+    private void WarnOnMissingPersistingClient()
+    {
+        if (this._agentOptions?.UseProvidedChatClientAsIs is not true)
+        {
+            return;
        }

-        return Task.CompletedTask;
+        if (this._agentOptions?.PersistChatHistoryAtEndOfRun is not true)
+        {
+            return;
+        }
+
+        var persistingClient = this.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+        if (persistingClient is null && this._logger.IsEnabled(LogLevel.Warning))
+        {
+            var loggingAgentName = this.GetLoggingAgentName();
+            this._logger.LogAgentChatClientMissingPersistingClient(
+                this.Id,
+                loggingAgentName);
+        }
    }

    private ChatHistoryProvider? ResolveChatHistoryProvider(ChatOptions? chatOptions, ChatClientAgentSession session)
@@ -69,4 +69,32 @@ internal static partial class ChatClientAgentLogMessages
        string chatHistoryProviderName,
        string agentId,
        string agentName);
+
+    /// <summary>
+    /// Logs a warning when <see cref="ChatClientAgentOptions.UseProvidedChatClientAsIs"/> is <see langword="true"/>
+    /// and <see cref="ChatClientAgentOptions.PersistChatHistoryAtEndOfRun"/> is <see langword="true"/>,
+    /// but no <see cref="ChatHistoryPersistingChatClient"/> is found in the custom chat client stack.
+    /// </summary>
+    [LoggerMessage(
+        Level = LogLevel.Warning,
+        Message = "Agent {AgentId}/{AgentName}: PersistChatHistoryAtEndOfRun is enabled with a custom chat client stack (UseProvidedChatClientAsIs), but no ChatHistoryPersistingChatClient was found in the pipeline. All messages will be persisted at the end of the run without marking. This setup is not supported with some other features, e.g. handoffs. Consider adding a ChatHistoryPersistingChatClient to the pipeline using the UseChatHistoryPersisting extension method.")]
+    public static partial void LogAgentChatClientMissingPersistingClient(
+        this ILogger logger,
+        string agentId,
+        string agentName);
+
+    /// <summary>
+    /// Logs a warning when per-service-call persistence falls back to end-of-run persistence
+    /// because the run involves background responses (continuation token resumption or
+    /// <c>AllowBackgroundResponses</c>). Per-service-call persistence is
+    /// unreliable in these scenarios because the caller may stop consuming the stream before
+    /// the decorator's post-stream persistence code can execute.
+    /// </summary>
+    [LoggerMessage(
+        Level = LogLevel.Warning,
+        Message = "Agent {AgentId}/{AgentName}: Per-service-call persistence is falling back to end-of-run persistence because the run involves background responses. Messages will be marked during the run and persisted at the end.")]
+    public static partial void LogAgentChatClientBackgroundResponseFallback(
+        this ILogger logger,
+        string agentId,
+        string agentName);
 }
@@ -1,7 +1,9 @@
 // Copyright (c) Microsoft. All rights reserved.

 using System.Collections.Generic;
+using System.Diagnostics.CodeAnalysis;
 using Microsoft.Extensions.AI;
+using Microsoft.Shared.DiagnosticIds;

 namespace Microsoft.Agents.AI;

@@ -89,6 +91,56 @@ public sealed class ChatClientAgentOptions
    /// </value>
    public bool ThrowOnChatHistoryProviderConflict { get; set; } = true;

+    /// <summary>
+    /// Gets or sets a value indicating whether to persist chat history only at the end of the full agent run
+    /// rather than after each individual service call.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// By default, <see cref="ChatClientAgent"/> persists request and response messages either via
+    /// a <see cref="ChatHistoryProvider"/>, or the underlying AI service's chat history storage.
+    /// Persistence is done immediately after each call to the AI service within the function invocation loop.
+    /// When storing in the underlying AI service, the session's <see cref="ChatClientAgentSession.ConversationId"/>
+    /// is also updated after each service call, keeping it in sync with the service-side conversation state.
+    /// </para>
+    /// <para>
+    /// Setting this property to <see langword="true"/> causes messages to be marked during the function
+    /// invocation loop but persisted only at the end of the full agent run, providing atomic run semantics.
+    /// Updating the <see cref="ChatClientAgentSession.ConversationId"/> is likewise deferred and
+    /// updated only at the end of the run, consistent with atomic run semantics.
+    /// A <see cref="ChatHistoryPersistingChatClient"/> decorator is inserted into the chat client pipeline
+    /// in mark-only mode, and the <see cref="ChatClientAgent"/> persists only the marked messages at the
+    /// end of the run.
+    /// </para>
+    /// <para>
+    /// When this option is <see langword="false"/> (the default), the <see cref="ChatHistoryPersistingChatClient"/>
+    /// decorator persists messages and updates the <see cref="ChatClientAgentSession.ConversationId"/>
+    /// immediately after each service call. This may leave chat history in a state where
+    /// <see cref="FunctionResultContent"/> is required to start a new run if the last successful service
+    /// call returned <see cref="FunctionCallContent"/>.
+    /// </para>
+    /// <para>
+    /// This option has no effect when <see cref="UseProvidedChatClientAsIs"/> is <see langword="true"/>.
+    /// When using a custom chat client stack, you can add a <see cref="ChatHistoryPersistingChatClient"/>
+    /// manually via the <see cref="ChatClientBuilderExtensions.UseChatHistoryPersisting"/>
+    /// extension method.
+    /// </para>
+    /// <para>
+    /// Note that when using single threaded service stored chat history, like OpenAI Conversations,
+    /// there is only one id, so even if the conversation id is not updated after each service call,
+    /// the chat history will still contain intermediate messages. Setting this property to <see langword="true"/>
+    /// in this case will therefore have no real effect. Setting this property to <see langword="true"/> when using
+    /// OpenAI Responses with response ids on the other hand, allows atomic run semantics, since
+    /// each service request produces a new response id, and if the run fails mid-loop, the session will
+    /// still contain the pre-run respnose id, allowing the next run to start with a clean slate.
+    /// </para>
+    /// </remarks>
+    /// <value>
+    /// Default is <see langword="false"/>.
+    /// </value>
+    [Experimental(DiagnosticIds.Experiments.AgentsAIExperiments)]
+    public bool PersistChatHistoryAtEndOfRun { get; set; }
+
    /// <summary>
    /// Creates a new instance of <see cref="ChatClientAgentOptions"/> with the same values as this instance.
    /// </summary>
@@ -105,5 +157,6 @@ public sealed class ChatClientAgentOptions
            ClearOnChatHistoryProviderConflict = this.ClearOnChatHistoryProviderConflict,
            WarnOnChatHistoryProviderConflict = this.WarnOnChatHistoryProviderConflict,
            ThrowOnChatHistoryProviderConflict = this.ThrowOnChatHistoryProviderConflict,
+            PersistChatHistoryAtEndOfRun = this.PersistChatHistoryAtEndOfRun,
        };
 }
@@ -2,8 +2,10 @@

 using System;
 using System.Collections.Generic;
+using System.Diagnostics.CodeAnalysis;
 using Microsoft.Agents.AI;
 using Microsoft.Extensions.Logging;
+using Microsoft.Shared.DiagnosticIds;
 using Microsoft.Shared.Diagnostics;

 namespace Microsoft.Extensions.AI;
@@ -82,4 +84,46 @@ public static class ChatClientBuilderExtensions
            options: options,
            loggerFactory: loggerFactory,
            services: services);
+
+    /// <summary>
+    /// Adds a <see cref="ChatHistoryPersistingChatClient"/> to the chat client pipeline.
+    /// </summary>
+    /// <remarks>
+    /// <para>
+    /// This decorator should be positioned between the <see cref="FunctionInvokingChatClient"/> and the leaf
+    /// <see cref="IChatClient"/> in the pipeline. It intercepts service calls to either persist messages
+    /// immediately or mark them for later persistence, depending on the <paramref name="markOnly"/> parameter.
+    /// </para>
+    /// <para>
+    /// If <paramref name="markOnly"/> is set to <see langword="true"/>, the <see cref="ChatClientAgent"/>
+    /// should be configured with <see cref="ChatClientAgentOptions.PersistChatHistoryAtEndOfRun"/> set to <see langword="true"/>
+    /// as without this combination, messages will never be persisted when using a <see cref="ChatHistoryProvider"/> for
+    /// chat history persistence.
+    /// </para>
+    /// <para>
+    /// This extension method is intended for use with custom chat client stacks when
+    /// <see cref="ChatClientAgentOptions.UseProvidedChatClientAsIs"/> is <see langword="true"/>.
+    /// When <see cref="ChatClientAgentOptions.UseProvidedChatClientAsIs"/> is <see langword="false"/> (the default),
+    /// the <see cref="ChatClientAgent"/> automatically injects this decorator.
+    /// </para>
+    /// <para>
+    /// This decorator only works within the context of a running <see cref="ChatClientAgent"/> and will throw an
+    /// exception if used in any other stack.
+    /// </para>
+    /// </remarks>
+    /// <param name="builder">The <see cref="ChatClientBuilder"/> to add the decorator to.</param>
+    /// <param name="markOnly">
+    /// When <see langword="true"/>, messages are marked with metadata but not persisted immediately,
+    /// and the session's <see cref="ChatClientAgentSession.ConversationId"/> is not updated.
+    /// The <see cref="ChatClientAgent"/> will persist only the marked messages and update the
+    /// conversation ID at the end of the run.
+    /// When <see langword="false"/> (the default), messages are persisted and the conversation ID
+    /// is updated immediately after each service call.
+    /// </param>
+    /// <returns>The <paramref name="builder"/> for chaining.</returns>
+    [Experimental(DiagnosticIds.Experiments.AgentsAIExperiments)]
+    public static ChatClientBuilder UseChatHistoryPersisting(this ChatClientBuilder builder, bool markOnly = false)
+    {
+        return builder.Use(innerClient => new ChatHistoryPersistingChatClient(innerClient, markOnly));
+    }
 }
@@ -63,6 +63,15 @@ public static class ChatClientExtensions
            });
        }

+        // ChatHistoryPersistingChatClient is registered after FunctionInvokingChatClient so that it sits
+        // between FIC and the leaf client. ChatClientBuilder.Build applies factories in reverse order,
+        // making the first Use() call outermost. By adding our decorator second, the resulting pipeline is:
+        //   FunctionInvokingChatClient → ChatHistoryPersistingChatClient → leaf IChatClient
+        // This allows the decorator to persist messages after each individual service call within
+        // FIC's function invocation loop, or to mark them for later persistence at the end of the run.
+        bool markOnly = options?.PersistChatHistoryAtEndOfRun is true;
+        chatBuilder.Use(innerClient => new ChatHistoryPersistingChatClient(innerClient, markOnly));
+
        var agentChatClient = chatBuilder.Build(services);

        if (options?.ChatOptions?.Tools is { Count: > 0 })
@@ -0,0 +1,313 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Runtime.CompilerServices;
+using System.Threading;
+using System.Threading.Tasks;
+using Microsoft.Extensions.AI;
+
+namespace Microsoft.Agents.AI;
+
+/// <summary>
+/// A delegating chat client that notifies <see cref="ChatHistoryProvider"/> and <see cref="AIContextProvider"/>
+/// instances of request and response messages after each individual call to the inner chat client,
+/// or marks messages for later persistence depending on the configured mode.
+/// </summary>
+/// <remarks>
+/// <para>
+/// This decorator is intended to operate between the <see cref="FunctionInvokingChatClient"/> and the leaf
+/// <see cref="IChatClient"/> in a <see cref="ChatClientAgent"/> pipeline.
+/// </para>
+/// <para>
+/// In persist mode (the default), it ensures that providers are notified and the session's
+/// <see cref="ChatClientAgentSession.ConversationId"/> is updated after each service call, so that
+/// intermediate messages (e.g., tool calls and results) are saved even if the process is interrupted
+/// mid-loop.
+/// </para>
+/// <para>
+/// In mark-only mode (<see cref="MarkOnly"/> is <see langword="true"/>), it marks messages with metadata
+/// but does not notify providers or update the <see cref="ChatClientAgentSession.ConversationId"/>.
+/// Both are deferred to the <see cref="ChatClientAgent"/> at the end of the run, providing atomic
+/// run semantics.
+/// </para>
+/// <para>
+/// This chat client must be used within the context of a running <see cref="ChatClientAgent"/>. It retrieves the
+/// current agent and session from <see cref="AIAgent.CurrentRunContext"/>, which is set automatically when an agent's
+/// <see cref="AIAgent.RunAsync(IEnumerable{ChatMessage}, AgentSession?, AgentRunOptions?, CancellationToken)"/> or
+/// <see cref="AIAgent.RunStreamingAsync(IEnumerable{ChatMessage}, AgentSession?, AgentRunOptions?, CancellationToken)"/>
+/// method is called. The <see cref="ChatClientAgent"/> ensures the run context always contains a resolved session,
+/// even when the caller passes null. An <see cref="InvalidOperationException"/> is thrown if no run context is
+/// available or if the agent is not a <see cref="ChatClientAgent"/>.
+/// </para>
+/// </remarks>
+internal sealed class ChatHistoryPersistingChatClient : DelegatingChatClient
+{
+    /// <summary>
+    /// The key used in <see cref="ChatMessage.AdditionalProperties"/> and <see cref="AIContent.AdditionalProperties"/>
+    /// to mark messages and their content as already persisted to chat history.
+    /// </summary>
+    internal const string PersistedMarkerKey = "_chatHistoryPersisted";
+
+    /// <summary>
+    /// Initializes a new instance of the <see cref="ChatHistoryPersistingChatClient"/> class.
+    /// </summary>
+    /// <param name="innerClient">The underlying chat client that will handle the core operations.</param>
+    /// <param name="markOnly">
+    /// When <see langword="true"/>, messages are marked with metadata but not persisted immediately,
+    /// and the session's <see cref="ChatClientAgentSession.ConversationId"/> is not updated.
+    /// The <see cref="ChatClientAgent"/> will persist only the marked messages and update the
+    /// conversation ID at the end of the run.
+    /// When <see langword="false"/> (the default), messages are persisted and the conversation ID
+    /// is updated immediately after each service call.
+    /// </param>
+    public ChatHistoryPersistingChatClient(IChatClient innerClient, bool markOnly = false)
+        : base(innerClient)
+    {
+        this.MarkOnly = markOnly;
+    }
+
+    /// <summary>
+    /// Gets a value indicating whether this decorator is in mark-only mode.
+    /// </summary>
+    /// <remarks>
+    /// When <see langword="true"/>, messages are marked with metadata but not persisted immediately,
+    /// and the session's <see cref="ChatClientAgentSession.ConversationId"/> is not updated.
+    /// Both are deferred to the <see cref="ChatClientAgent"/> at the end of the run.
+    /// When <see langword="false"/>, messages are persisted and the conversation ID is updated
+    /// after each service call.
+    /// </remarks>
+    public bool MarkOnly { get; }
+
+    /// <inheritdoc/>
+    public override async Task<ChatResponse> GetResponseAsync(
+        IEnumerable<ChatMessage> messages,
+        ChatOptions? options = null,
+        CancellationToken cancellationToken = default)
+    {
+        var (agent, session) = GetRequiredAgentAndSession();
+
+        ChatResponse response;
+        try
+        {
+            response = await base.GetResponseAsync(messages, options, cancellationToken).ConfigureAwait(false);
+        }
+        catch (Exception ex)
+        {
+            var newRequestMessagesOnFailure = GetNewRequestMessages(messages);
+            await agent.NotifyProvidersOfFailureAsync(session, ex, newRequestMessagesOnFailure, options, cancellationToken).ConfigureAwait(false);
+            throw;
+        }
+
+        var newRequestMessages = GetNewRequestMessages(messages);
+
+        if (this.ShouldDeferPersistence(options))
+        {
+            // In mark-only mode or when resuming from a continuation token, just mark messages
+            // for later persistence by ChatClientAgent. Conversation ID and provider notification
+            // are deferred to end-of-run. For continuation tokens, the end-of-run handler needs
+            // to send the combined data from both the previous and current runs.
+            MarkAsPersisted(newRequestMessages);
+            MarkAsPersisted(response.Messages);
+        }
+        else
+        {
+            // In persist mode, persist immediately and update conversation ID.
+            agent.UpdateSessionConversationId(session, response.ConversationId, cancellationToken);
+            await agent.NotifyProvidersOfNewMessagesAsync(session, newRequestMessages, response.Messages, options, cancellationToken).ConfigureAwait(false);
+            MarkAsPersisted(newRequestMessages);
+            MarkAsPersisted(response.Messages);
+        }
+
+        return response;
+    }
+
+    /// <inheritdoc/>
+    public override async IAsyncEnumerable<ChatResponseUpdate> GetStreamingResponseAsync(
+        IEnumerable<ChatMessage> messages,
+        ChatOptions? options = null,
+        [EnumeratorCancellation] CancellationToken cancellationToken = default)
+    {
+        var (agent, session) = GetRequiredAgentAndSession();
+
+        List<ChatResponseUpdate> responseUpdates = [];
+
+        IAsyncEnumerator<ChatResponseUpdate> enumerator;
+        try
+        {
+            enumerator = base.GetStreamingResponseAsync(messages, options, cancellationToken).GetAsyncEnumerator(cancellationToken);
+        }
+        catch (Exception ex)
+        {
+            var newRequestMessagesOnFailure = GetNewRequestMessages(messages);
+            await agent.NotifyProvidersOfFailureAsync(session, ex, newRequestMessagesOnFailure, options, cancellationToken).ConfigureAwait(false);
+            throw;
+        }
+
+        bool hasUpdates;
+        try
+        {
+            hasUpdates = await enumerator.MoveNextAsync().ConfigureAwait(false);
+        }
+        catch (Exception ex)
+        {
+            var newRequestMessagesOnFailure = GetNewRequestMessages(messages);
+            await agent.NotifyProvidersOfFailureAsync(session, ex, newRequestMessagesOnFailure, options, cancellationToken).ConfigureAwait(false);
+            throw;
+        }
+
+        while (hasUpdates)
+        {
+            var update = enumerator.Current;
+            responseUpdates.Add(update);
+            yield return update;
+
+            try
+            {
+                hasUpdates = await enumerator.MoveNextAsync().ConfigureAwait(false);
+            }
+            catch (Exception ex)
+            {
+                var newRequestMessagesOnFailure = GetNewRequestMessages(messages);
+                await agent.NotifyProvidersOfFailureAsync(session, ex, newRequestMessagesOnFailure, options, cancellationToken).ConfigureAwait(false);
+                throw;
+            }
+        }
+
+        var chatResponse = responseUpdates.ToChatResponse();
+        var newRequestMessages = GetNewRequestMessages(messages);
+
+        if (this.ShouldDeferPersistence(options))
+        {
+            // In mark-only mode or when resuming from a continuation token, just mark messages
+            // for later persistence by ChatClientAgent. Conversation ID and provider notification
+            // are deferred to end-of-run. For continuation tokens, the end-of-run handler needs
+            // to send the combined data from both the previous and current runs.
+            MarkAsPersisted(newRequestMessages);
+            MarkAsPersisted(chatResponse.Messages);
+        }
+        else
+        {
+            // In persist mode, persist immediately and update conversation ID.
+            agent.UpdateSessionConversationId(session, chatResponse.ConversationId, cancellationToken);
+            await agent.NotifyProvidersOfNewMessagesAsync(session, newRequestMessages, chatResponse.Messages, options, cancellationToken).ConfigureAwait(false);
+            MarkAsPersisted(newRequestMessages);
+            MarkAsPersisted(chatResponse.Messages);
+        }
+    }
+
+    /// <summary>
+    /// Gets the current <see cref="ChatClientAgent"/> and <see cref="ChatClientAgentSession"/> from the run context.
+    /// </summary>
+    private static (ChatClientAgent Agent, ChatClientAgentSession Session) GetRequiredAgentAndSession()
+    {
+        var runContext = AIAgent.CurrentRunContext
+            ?? throw new InvalidOperationException(
+                $"{nameof(ChatHistoryPersistingChatClient)} can only be used within the context of a running AIAgent. " +
+                "Ensure that the chat client is being invoked as part of an AIAgent.RunAsync or AIAgent.RunStreamingAsync call.");
+
+        var chatClientAgent = runContext.Agent.GetService<ChatClientAgent>()
+            ?? throw new InvalidOperationException(
+                $"{nameof(ChatHistoryPersistingChatClient)} can only be used with a {nameof(ChatClientAgent)}. " +
+                $"The current agent is of type '{runContext.Agent.GetType().Name}'.");
+
+        if (runContext.Session is not ChatClientAgentSession chatClientAgentSession)
+        {
+            throw new InvalidOperationException(
+                $"{nameof(ChatHistoryPersistingChatClient)} requires a {nameof(ChatClientAgentSession)}. " +
+                $"The current session is of type '{runContext.Session?.GetType().Name ?? "null"}'.");
+        }
+
+        return (chatClientAgent, chatClientAgentSession);
+    }
+
+    /// <summary>
+    /// Determines whether persistence should be deferred to end-of-run instead of happening immediately.
+    /// </summary>
+    /// <returns>
+    /// <see langword="true"/> when in <see cref="MarkOnly"/> mode, when the call is resuming from
+    /// a continuation token (since the end-of-run handler needs to combine data from the previous
+    /// and current runs), or when background responses are allowed (since the caller may stop
+    /// consuming the stream mid-run, preventing the post-stream persistence code from executing).
+    /// </returns>
+    private bool ShouldDeferPersistence(ChatOptions? options)
+    {
+        return this.MarkOnly || options?.ContinuationToken is not null || options?.AllowBackgroundResponses is true;
+    }
+
+    /// <summary>
+    /// Returns only the request messages that have not yet been persisted to chat history.
+    /// </summary>
+    /// <remarks>
+    /// A message is considered already persisted if any of the following is true:
+    /// <list type="bullet">
+    /// <item>It has the <see cref="PersistedMarkerKey"/> in its <see cref="ChatMessage.AdditionalProperties"/>.</item>
+    /// <item>It has an <see cref="AgentRequestMessageSourceType"/> of <see cref="AgentRequestMessageSourceType.ChatHistory"/>
+    /// (indicating it was loaded from chat history and does not need to be re-persisted).</item>
+    /// <item>It has <see cref="ChatMessage.Contents"/> and all of its <see cref="AIContent"/> items have the
+    /// <see cref="PersistedMarkerKey"/> in their <see cref="AIContent.AdditionalProperties"/>. This handles the
+    /// streaming case where <see cref="FunctionInvokingChatClient"/> reconstructs <see cref="ChatMessage"/> objects
+    /// independently via <c>ToChatResponse()</c>, producing different object references that share the same
+    /// underlying <see cref="AIContent"/> instances.</item>
+    /// </list>
+    /// </remarks>
+    /// <returns>A list of request messages that have not yet been persisted.</returns>
+    /// <param name="messages">The full set of request messages to filter.</param>
+    private static List<ChatMessage> GetNewRequestMessages(IEnumerable<ChatMessage> messages)
+    {
+        return messages.Where(m => !IsAlreadyPersisted(m)).ToList();
+    }
+
+    /// <summary>
+    /// Determines whether a message has already been persisted to chat history by this decorator.
+    /// </summary>
+    private static bool IsAlreadyPersisted(ChatMessage message)
+    {
+        if (message.AdditionalProperties?.TryGetValue(PersistedMarkerKey, out var value) == true && value is true)
+        {
+            return true;
+        }
+
+        if (message.GetAgentRequestMessageSourceType() == AgentRequestMessageSourceType.ChatHistory)
+        {
+            return true;
+        }
+
+        // In streaming mode, FunctionInvokingChatClient reconstructs ChatMessage objects via ToChatResponse()
+        // independently, producing different ChatMessage instances. However, the underlying AIContent objects
+        // (e.g., FunctionCallContent, FunctionResultContent) are shared references. Checking for markers on
+        // AIContent handles dedup in this case.
+        if (message.Contents.Count > 0 && message.Contents.All(c => c.AdditionalProperties?.TryGetValue(PersistedMarkerKey, out var value) == true && value is true))
+        {
+            return true;
+        }
+
+        return false;
+    }
+
+    /// <summary>
+    /// Marks the given messages as persisted by setting a marker on both the <see cref="ChatMessage"/>
+    /// and each of its <see cref="AIContent"/> items.
+    /// </summary>
+    /// <remarks>
+    /// Both levels are marked because <see cref="FunctionInvokingChatClient"/> may reconstruct
+    /// <see cref="ChatMessage"/> objects in streaming mode (losing the message-level marker),
+    /// but the <see cref="AIContent"/> references are shared and retain their markers.
+    /// </remarks>
+    /// <param name="messages">The messages to mark as persisted.</param>
+    private static void MarkAsPersisted(IEnumerable<ChatMessage> messages)
+    {
+        foreach (var message in messages)
+        {
+            message.AdditionalProperties ??= new();
+            message.AdditionalProperties[PersistedMarkerKey] = true;
+
+            foreach (var content in message.Contents)
+            {
+                content.AdditionalProperties ??= new();
+                content.AdditionalProperties[PersistedMarkerKey] = true;
+            }
+        }
+    }
+}
@@ -5,6 +5,8 @@ using System.Reflection;
 using System.Text;
 using Microsoft.Extensions.Configuration;
 using Microsoft.Extensions.Logging;
+using ModelContextProtocol.Client;
+using ModelContextProtocol.Protocol;
 namespace Microsoft.Agents.AI.Hosting.AzureFunctions.IntegrationTests;

 /// <summary>
@@ -235,6 +237,114 @@ public sealed class WorkflowSamplesValidation(ITestOutputHelper outputHelper) :
        });
    }

+    [Fact]
+    public async Task WorkflowMcpToolSampleValidationAsync()
+    {
+        string samplePath = Path.Combine(s_samplesPath, "04_WorkflowMcpTool");
+        await this.RunSampleTestAsync(samplePath, requiresOpenAI: false, async (logs) =>
+        {
+            // Connect to the MCP endpoint exposed by the Azure Functions host
+            IClientTransport clientTransport = new HttpClientTransport(new()
+            {
+                Endpoint = new Uri($"http://localhost:{AzureFunctionsPort}/runtime/webhooks/mcp")
+            });
+
+            await using McpClient mcpClient = await McpClient.CreateAsync(clientTransport);
+
+            // Verify both workflow tools are listed
+            IList<McpClientTool> tools = await mcpClient.ListToolsAsync();
+            this._outputHelper.WriteLine($"MCP tools found: {string.Join(", ", tools.Select(t => t.Name))}");
+
+            Assert.Single(tools, t => t.Name == "Translate");
+            Assert.Single(tools, t => t.Name == "OrderLookup");
+
+            // Invoke the Translate workflow via MCP tool (returns a string result)
+            this._outputHelper.WriteLine("Invoking MCP tool 'Translate'...");
+            CallToolResult translateResult = await mcpClient.CallToolAsync(
+                "Translate",
+                arguments: new Dictionary<string, object?> { { "input", "hello world" } });
+
+            Assert.NotEmpty(translateResult.Content);
+            string translateResponse = Assert.IsType<TextContentBlock>(translateResult.Content[0]).Text;
+            this._outputHelper.WriteLine($"Translate MCP tool response: {translateResponse}");
+            Assert.NotEmpty(translateResponse);
+            Assert.Contains("HELLO WORLD", translateResponse);
+
+            // Invoke the OrderLookup workflow via MCP tool (returns a POCO serialized as JSON)
+            this._outputHelper.WriteLine("Invoking MCP tool 'OrderLookup'...");
+            CallToolResult orderResult = await mcpClient.CallToolAsync(
+                "OrderLookup",
+                arguments: new Dictionary<string, object?> { { "input", "ORD-2025-42" } });
+
+            Assert.NotEmpty(orderResult.Content);
+            string orderResponse = Assert.IsType<TextContentBlock>(orderResult.Content[0]).Text;
+            this._outputHelper.WriteLine($"OrderLookup MCP tool response: {orderResponse}");
+            Assert.NotEmpty(orderResponse);
+            Assert.Contains("ORD-2025-42", orderResponse);
+
+            // Verify executor activities ran in the logs
+            lock (logs)
+            {
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] TranslateText:")), "TranslateText activity not found in logs.");
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] FormatOutput:")), "FormatOutput activity not found in logs.");
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] LookupOrder:")), "LookupOrder activity not found in logs.");
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] EnrichOrder:")), "EnrichOrder activity not found in logs.");
+            }
+        });
+    }
+
+    [Fact]
+    public async Task WorkflowAndAgentsSampleValidationAsync()
+    {
+        string samplePath = Path.Combine(s_samplesPath, "05_WorkflowAndAgents");
+        await this.RunSampleTestAsync(samplePath, requiresOpenAI: true, async (logs) =>
+        {
+            // Connect to the MCP endpoint exposed by the Azure Functions host
+            IClientTransport clientTransport = new HttpClientTransport(new()
+            {
+                Endpoint = new Uri($"http://localhost:{AzureFunctionsPort}/runtime/webhooks/mcp")
+            });
+
+            await using McpClient mcpClient = await McpClient.CreateAsync(clientTransport);
+
+            // Verify both the agent and workflow tools are listed
+            IList<McpClientTool> tools = await mcpClient.ListToolsAsync();
+            this._outputHelper.WriteLine($"MCP tools found: {string.Join(", ", tools.Select(t => t.Name))}");
+
+            Assert.Single(tools, t => t.Name == "Assistant");
+            Assert.Single(tools, t => t.Name == "Translate");
+
+            // Invoke the Translate workflow via MCP tool
+            this._outputHelper.WriteLine("Invoking MCP tool 'Translate'...");
+            CallToolResult translateResult = await mcpClient.CallToolAsync(
+                "Translate",
+                arguments: new Dictionary<string, object?> { { "input", "hello world" } });
+
+            Assert.NotEmpty(translateResult.Content);
+            string translateResponse = Assert.IsType<TextContentBlock>(translateResult.Content[0]).Text;
+            this._outputHelper.WriteLine($"Translate MCP tool response: {translateResponse}");
+            Assert.Contains("HELLO WORLD", translateResponse);
+
+            // Invoke the Assistant agent via MCP tool
+            this._outputHelper.WriteLine("Invoking MCP tool 'Assistant'...");
+            CallToolResult assistantResult = await mcpClient.CallToolAsync(
+                "Assistant",
+                arguments: new Dictionary<string, object?> { { "query", "What is 2 + 2?" } });
+
+            Assert.NotEmpty(assistantResult.Content);
+            string assistantResponse = Assert.IsType<TextContentBlock>(assistantResult.Content[0]).Text;
+            this._outputHelper.WriteLine($"Assistant MCP tool response: {assistantResponse}");
+            Assert.NotEmpty(assistantResponse);
+
+            // Verify workflow executor activities ran in the logs
+            lock (logs)
+            {
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] TranslateText:")), "TranslateText activity not found in logs.");
+                Assert.True(logs.Any(log => log.Message.Contains("[Activity] FormatOutput:")), "FormatOutput activity not found in logs.");
+            }
+        });
+    }
+
    [Fact]
    public async Task ConcurrentWorkflowSampleValidationAsync()
    {
@@ -148,6 +148,45 @@ public sealed class DurableAgentFunctionMetadataTransformerTests
        }
    }

+    [Fact]
+    public void Transform_SkipsAgents_WithoutExplicitOptions()
+    {
+        // Arrange: two agents in the dictionary, but only one has explicit FunctionsAgentOptions.
+        // This simulates a workflow-auto-registered agent (workflowAgent) alongside a standalone agent.
+        Dictionary<string, Func<IServiceProvider, AIAgent>> agents = new()
+        {
+            { "standaloneAgent", _ => new TestAgent("standaloneAgent", "Standalone agent") },
+            { "workflowAgent", _ => new TestAgent("workflowAgent", "Auto-registered by workflow") }
+        };
+
+        FunctionsAgentOptions standaloneOptions = new();
+        standaloneOptions.HttpTrigger.IsEnabled = true;
+
+        // Only standaloneAgent has explicit options; workflowAgent does not.
+        IFunctionsAgentOptionsProvider agentOptionsProvider = new FakeOptionsProvider(new Dictionary<string, FunctionsAgentOptions>
+        {
+            { "standaloneAgent", standaloneOptions }
+        });
+
+        List<IFunctionMetadata> metadataList = [];
+
+        DurableAgentFunctionMetadataTransformer transformer = new(
+            agents,
+            NullLogger<DurableAgentFunctionMetadataTransformer>.Instance,
+            new FakeServiceProvider(),
+            agentOptionsProvider);
+
+        // Act
+        transformer.Transform(metadataList);
+
+        // Assert: only standaloneAgent should have triggers (entity + http = 2).
+        // workflowAgent should be skipped entirely.
+        Assert.Equal(2, metadataList.Count);
+        Assert.Contains(metadataList, m => m.Name == "dafx-standaloneAgent");
+        Assert.Contains(metadataList, m => m.Name == "http-standaloneAgent");
+        Assert.DoesNotContain(metadataList, m => m.Name!.Contains("workflowAgent"));
+    }
+
    private static List<IFunctionMetadata> BuildFunctionMetadataList(int numberOfFunctions)
    {
        List<IFunctionMetadata> list = [];
@@ -0,0 +1,121 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using System.Text.Json;
+using Microsoft.Azure.Functions.Worker.Core.FunctionMetadata;
+
+namespace Microsoft.Agents.AI.Hosting.AzureFunctions.UnitTests;
+
+public sealed class FunctionMetadataFactoryTests
+{
+    [Fact]
+    public void CreateEntityTrigger_SetsCorrectNameAndBindings()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateEntityTrigger("myAgent");
+
+        Assert.Equal("dafx-myAgent", metadata.Name);
+        Assert.Equal("dotnet-isolated", metadata.Language);
+        Assert.Equal(BuiltInFunctions.RunAgentEntityFunctionEntryPoint, metadata.EntryPoint);
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Equal(2, metadata.RawBindings.Count);
+        Assert.Contains("entityTrigger", metadata.RawBindings[0]);
+        Assert.Contains("durableClient", metadata.RawBindings[1]);
+    }
+
+    [Fact]
+    public void CreateHttpTrigger_SetsCorrectNameRouteAndDefaults()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateHttpTrigger(
+            "myWorkflow", "workflows/myWorkflow/run", BuiltInFunctions.RunWorkflowOrchestrationHttpFunctionEntryPoint);
+
+        Assert.Equal("http-myWorkflow", metadata.Name);
+        Assert.Equal("dotnet-isolated", metadata.Language);
+        Assert.Equal(BuiltInFunctions.RunWorkflowOrchestrationHttpFunctionEntryPoint, metadata.EntryPoint);
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Equal(3, metadata.RawBindings.Count);
+        Assert.Contains("httpTrigger", metadata.RawBindings[0]);
+        Assert.Contains("workflows/myWorkflow/run", metadata.RawBindings[0]);
+        Assert.Contains("\"post\"", metadata.RawBindings[0]);
+        Assert.Contains("http", metadata.RawBindings[1]);
+        Assert.Contains("durableClient", metadata.RawBindings[2]);
+    }
+
+    [Fact]
+    public void CreateHttpTrigger_RespectsCustomMethods()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateHttpTrigger(
+            "status", "workflows/status/{runId}", BuiltInFunctions.GetWorkflowStatusHttpFunctionEntryPoint, methods: "\"get\"");
+
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Contains("\"get\"", metadata.RawBindings[0]);
+        Assert.DoesNotContain("\"post\"", metadata.RawBindings[0]);
+    }
+
+    [Fact]
+    public void CreateActivityTrigger_SetsCorrectNameAndBindings()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateActivityTrigger("dafx-MyExecutor");
+
+        Assert.Equal("dafx-MyExecutor", metadata.Name);
+        Assert.Equal("dotnet-isolated", metadata.Language);
+        Assert.Equal(BuiltInFunctions.InvokeWorkflowActivityFunctionEntryPoint, metadata.EntryPoint);
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Equal(2, metadata.RawBindings.Count);
+        Assert.Contains("activityTrigger", metadata.RawBindings[0]);
+        Assert.Contains("durableClient", metadata.RawBindings[1]);
+    }
+
+    [Fact]
+    public void CreateOrchestrationTrigger_SetsCorrectNameAndBindings()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateOrchestrationTrigger(
+            "dafx-MyWorkflow", BuiltInFunctions.RunWorkflowOrchestrationFunctionEntryPoint);
+
+        Assert.Equal("dafx-MyWorkflow", metadata.Name);
+        Assert.Equal("dotnet-isolated", metadata.Language);
+        Assert.Equal(BuiltInFunctions.RunWorkflowOrchestrationFunctionEntryPoint, metadata.EntryPoint);
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Single(metadata.RawBindings);
+        Assert.Contains("orchestrationTrigger", metadata.RawBindings[0]);
+    }
+
+    [Fact]
+    public void CreateWorkflowMcpToolTrigger_SetsCorrectNameAndBindings()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateWorkflowMcpToolTrigger("Translate", "Translate text");
+
+        Assert.Equal("mcptool-Translate", metadata.Name);
+        Assert.Equal("dotnet-isolated", metadata.Language);
+        Assert.Equal(BuiltInFunctions.RunWorkflowMcpToolFunctionEntryPoint, metadata.EntryPoint);
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Equal(3, metadata.RawBindings.Count);
+
+        // Verify all bindings are valid JSON
+        foreach (string binding in metadata.RawBindings)
+        {
+            JsonDocument.Parse(binding);
+        }
+
+        // mcpToolTrigger binding
+        Assert.Contains("mcpToolTrigger", metadata.RawBindings[0]);
+        Assert.Contains("\"toolName\":\"Translate\"", metadata.RawBindings[0]);
+        Assert.Contains("\"description\":\"Translate text\"", metadata.RawBindings[0]);
+        Assert.Contains("toolProperties", metadata.RawBindings[0]);
+
+        // mcpToolProperty binding for input
+        Assert.Contains("mcpToolProperty", metadata.RawBindings[1]);
+        Assert.Contains("\"propertyName\":\"input\"", metadata.RawBindings[1]);
+        Assert.Contains("\"isRequired\":true", metadata.RawBindings[1]);
+
+        // durableClient binding
+        Assert.Contains("durableClient", metadata.RawBindings[2]);
+    }
+
+    [Fact]
+    public void CreateWorkflowMcpToolTrigger_UsesDefaultDescription_WhenNull()
+    {
+        DefaultFunctionMetadata metadata = FunctionMetadataFactory.CreateWorkflowMcpToolTrigger("MyWorkflow", description: null);
+
+        Assert.NotNull(metadata.RawBindings);
+        Assert.Contains("Run the MyWorkflow workflow", metadata.RawBindings[0]);
+    }
+}
@@ -250,6 +250,129 @@ public class AIContextProviderChatClientTests

    #endregion

+    #region Shared Options Tests
+
+    [Fact]
+    public async Task GetResponseAsync_SharedOptions_ProviderToolsDoNotAccumulateAcrossCallsAsync()
+    {
+        // Arrange: track tool count seen by the inner client on each call
+        var toolCountsSeenByInner = new List<int>();
+
+        var innerClient = CreateMockChatClient(
+            onGetResponse: (_, options, _) =>
+            {
+                toolCountsSeenByInner.Add(options?.Tools?.Count ?? 0);
+                return Task.FromResult(new ChatResponse([new ChatMessage(ChatRole.Assistant, "Response")]));
+            });
+
+        var provider = new TestAIContextProvider("key1", provideTools: [new TestAITool()]);
+        var chatClient = new AIContextProviderChatClient(innerClient, [provider]);
+
+        var sharedOptions = new ChatOptions
+        {
+            Tools = new List<AITool> { new TestAITool() }
+        };
+
+        // Act: make 3 calls reusing the same ChatOptions
+        for (int i = 0; i < 3; i++)
+        {
+            await RunWithAgentContextAsync(chatClient, sharedOptions);
+        }
+
+        // Assert: each call should see exactly 2 tools (1 baseline + 1 injected)
+        Assert.Equal(3, toolCountsSeenByInner.Count);
+        Assert.All(toolCountsSeenByInner, count => Assert.Equal(2, count));
+    }
+
+    [Fact]
+    public async Task GetResponseAsync_SharedOptions_OriginalToolsNotMutatedAsync()
+    {
+        // Arrange
+        var innerClient = CreateMockChatClient(
+            onGetResponse: (_, _, _) =>
+                Task.FromResult(new ChatResponse([new ChatMessage(ChatRole.Assistant, "Response")])));
+
+        var provider = new TestAIContextProvider("key1", provideTools: [new TestAITool()]);
+        var chatClient = new AIContextProviderChatClient(innerClient, [provider]);
+
+        var baselineTool = new TestAITool();
+        var originalTools = new List<AITool> { baselineTool };
+        var sharedOptions = new ChatOptions
+        {
+            Tools = originalTools
+        };
+
+        // Act
+        await RunWithAgentContextAsync(chatClient, sharedOptions);
+
+        // Assert: the original list should still contain only the baseline tool
+        Assert.Single(originalTools);
+        Assert.Same(baselineTool, originalTools[0]);
+        Assert.Same(originalTools, sharedOptions.Tools);
+        Assert.Same(baselineTool, originalTools[0]);
+    }
+
+    [Fact]
+    public async Task GetStreamingResponseAsync_SharedOptions_ProviderToolsDoNotAccumulateAcrossCallsAsync()
+    {
+        // Arrange
+        var toolCountsSeenByInner = new List<int>();
+
+        var innerClient = CreateMockStreamingChatClient(
+            onGetStreamingResponse: (_, options, _) =>
+            {
+                toolCountsSeenByInner.Add(options?.Tools?.Count ?? 0);
+                return ToAsyncEnumerableAsync(
+                    new ChatResponseUpdate(ChatRole.Assistant, "Response"));
+            });
+
+        var provider = new TestAIContextProvider("key1", provideTools: [new TestAITool()]);
+        var chatClient = new AIContextProviderChatClient(innerClient, [provider]);
+
+        var sharedOptions = new ChatOptions
+        {
+            Tools = new List<AITool> { new TestAITool() }
+        };
+
+        // Act: make 3 streaming calls reusing the same ChatOptions
+        for (int i = 0; i < 3; i++)
+        {
+            await RunStreamingWithAgentContextAsync(chatClient, [], sharedOptions);
+        }
+
+        // Assert: each call should see exactly 2 tools (1 baseline + 1 injected)
+        Assert.Equal(3, toolCountsSeenByInner.Count);
+        Assert.All(toolCountsSeenByInner, count => Assert.Equal(2, count));
+    }
+
+    [Fact]
+    public async Task GetStreamingResponseAsync_SharedOptions_OriginalToolsNotMutatedAsync()
+    {
+        // Arrange
+        var innerClient = CreateMockStreamingChatClient(
+            onGetStreamingResponse: (_, _, _) => ToAsyncEnumerableAsync(
+                new ChatResponseUpdate(ChatRole.Assistant, "Response")));
+
+        var provider = new TestAIContextProvider("key1", provideTools: [new TestAITool()]);
+        var chatClient = new AIContextProviderChatClient(innerClient, [provider]);
+
+        var baselineTool = new TestAITool();
+        var originalTools = new List<AITool> { baselineTool };
+        var sharedOptions = new ChatOptions
+        {
+            Tools = originalTools
+        };
+
+        // Act
+        await RunStreamingWithAgentContextAsync(chatClient, [], sharedOptions);
+
+        // Assert: the original list should still contain only the baseline tool
+        Assert.Single(originalTools);
+        Assert.Same(baselineTool, originalTools[0]);
+    }
+
+    #endregion
+
    #region Builder Extension Tests

    [Fact]
@@ -341,6 +464,44 @@ public class AIContextProviderChatClientTests
        await agent.RunAsync([new ChatMessage(ChatRole.User, "Hello")], s_mockSession);
    }

+    /// <summary>
+    /// Runs a chat client within an agent context with the specified options.
+    /// </summary>
+    private static async Task RunWithAgentContextAsync(AIContextProviderChatClient chatClient, ChatOptions options)
+    {
+        var agent = new TestAIAgent
+        {
+            RunAsyncFunc = async (messages, session, agentOptions, ct) =>
+            {
+                var response = await chatClient.GetResponseAsync(messages, options, ct);
+                return new AgentResponse(response);
+            }
+        };
+
+        await agent.RunAsync([new ChatMessage(ChatRole.User, "Hello")], s_mockSession);
+    }
+
+    /// <summary>
+    /// Runs a streaming chat client within an agent context with the specified options.
+    /// </summary>
+    private static async Task RunStreamingWithAgentContextAsync(AIContextProviderChatClient chatClient, List<ChatResponseUpdate> updates, ChatOptions options)
+    {
+        var agent = new TestAIAgent
+        {
+            RunAsyncFunc = async (messages, session, agentOptions, ct) =>
+            {
+                await foreach (var update in chatClient.GetStreamingResponseAsync(messages, options, ct))
+                {
+                    updates.Add(update);
+                }
+
+                return new AgentResponse([new ChatMessage(ChatRole.Assistant, "done")]);
+            }
+        };
+
+        await agent.RunAsync([new ChatMessage(ChatRole.User, "Hello")], s_mockSession);
+    }
+
    private static IChatClient CreateMockChatClient(
        Func<IEnumerable<ChatMessage>, ChatOptions?, CancellationToken, Task<ChatResponse>> onGetResponse)
    {
@@ -0,0 +1,766 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using System;
+using System.Collections.Generic;
+using System.Linq;
+using System.Threading;
+using System.Threading.Tasks;
+using Microsoft.Extensions.AI;
+using Microsoft.Extensions.DependencyInjection;
+using Moq;
+using Moq.Protected;
+
+namespace Microsoft.Agents.AI.UnitTests;
+
+/// <summary>
+/// Contains unit tests for the <see cref="ChatHistoryPersistingChatClient"/> decorator,
+/// verifying that it persists messages via the <see cref="ChatHistoryProvider"/> after each
+/// individual service call by default, or marks messages for end-of-run persistence when the
+/// <see cref="ChatClientAgentOptions.PersistChatHistoryAtEndOfRun"/> option is enabled.
+/// </summary>
+public class ChatHistoryPersistingChatClientTests
+{
+    /// <summary>
+    /// Verifies that by default (PersistChatHistoryAtEndOfRun is false),
+    /// the ChatHistoryProvider receives messages after a successful non-streaming call.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_PersistsMessagesPerServiceCall_ByDefaultAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")]));
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert — InvokedCoreAsync should be called by the decorator (per service call)
+        mockChatHistoryProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<ChatHistoryProvider.InvokedContext>(x =>
+                    x.RequestMessages.Any(m => m.Text == "test") &&
+                    x.ResponseMessages!.Any(m => m.Text == "response")),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default),
+    /// the ChatHistoryProvider receives messages at the end of the run.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_PersistsMessagesAtEndOfRun_WhenOptionEnabledAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")]));
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = true,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert — InvokedCoreAsync should be called once by the agent (end of run)
+        mockChatHistoryProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<ChatHistoryProvider.InvokedContext>(x =>
+                    x.RequestMessages.Any(m => m.Text == "test") &&
+                    x.ResponseMessages!.Any(m => m.Text == "response")),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default) and the service call fails,
+    /// the ChatHistoryProvider is notified with the exception.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_NotifiesProviderOfFailure_WhenPerServiceCallPersistenceActiveAsync()
+    {
+        // Arrange
+        var expectedException = new InvalidOperationException("Service failed");
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ThrowsAsync(expectedException);
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await Assert.ThrowsAsync<InvalidOperationException>(() => agent.RunAsync([new(ChatRole.User, "test")], session));
+
+        // Assert — the decorator should have notified the provider of the failure
+        mockChatHistoryProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<ChatHistoryProvider.InvokedContext>(x =>
+                    x.InvokeException != null &&
+                    x.InvokeException.Message == "Service failed"),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that the decorator is injected in persist mode by default
+    /// and can be discovered via GetService.
+    /// </summary>
+    [Fact]
+    public void ChatClient_ContainsDecorator_InPersistMode_ByDefault()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+
+        // Act
+        ChatClientAgent agent = new(mockService.Object, options: new());
+
+        // Assert
+        var decorator = agent.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+        Assert.NotNull(decorator);
+        Assert.False(decorator.MarkOnly);
+    }
+
+    /// <summary>
+    /// Verifies that the decorator is injected in mark-only mode when PersistChatHistoryAtEndOfRun is true.
+    /// </summary>
+    [Fact]
+    public void ChatClient_ContainsDecorator_InMarkOnlyMode_WhenPersistAtEndOfRun()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+
+        // Act
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            PersistChatHistoryAtEndOfRun = true,
+        });
+
+        // Assert
+        var decorator = agent.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+        Assert.NotNull(decorator);
+        Assert.True(decorator.MarkOnly);
+    }
+
+    /// <summary>
+    /// Verifies that the decorator is NOT injected when UseProvidedChatClientAsIs is true.
+    /// </summary>
+    [Fact]
+    public void ChatClient_DoesNotContainDecorator_WhenUseProvidedChatClientAsIs()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+
+        // Act
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            UseProvidedChatClientAsIs = true,
+        });
+
+        // Assert
+        var decorator = agent.ChatClient.GetService<ChatHistoryPersistingChatClient>();
+        Assert.Null(decorator);
+    }
+
+    /// <summary>
+    /// Verifies that the PersistChatHistoryAtEndOfRun option is included in Clone().
+    /// </summary>
+    [Fact]
+    public void ChatClientAgentOptions_Clone_IncludesPersistChatHistoryAtEndOfRun()
+    {
+        // Arrange
+        var options = new ChatClientAgentOptions
+        {
+            PersistChatHistoryAtEndOfRun = true,
+        };
+
+        // Act
+        var cloned = options.Clone();
+
+        // Assert
+        Assert.True(cloned.PersistChatHistoryAtEndOfRun);
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default) and the service call
+    /// involves a function invocation loop, the ChatHistoryProvider is called after each individual
+    /// service call (not just once at the end).
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_PersistsPerServiceCall_DuringFunctionInvocationLoopAsync()
+    {
+        // Arrange
+        int serviceCallCount = 0;
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>()))
+            .Returns(() =>
+            {
+                serviceCallCount++;
+                if (serviceCallCount == 1)
+                {
+                    // First call returns a tool call
+                    return Task.FromResult(new ChatResponse([new(ChatRole.Assistant, [new FunctionCallContent("call1", "myTool", new Dictionary<string, object?>())])]));
+                }
+
+                // Second call returns a final response
+                return Task.FromResult(new ChatResponse([new(ChatRole.Assistant, "final response")]));
+            });
+
+        var invokedContexts = new List<ChatHistoryProvider.InvokedContext>();
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Callback((ChatHistoryProvider.InvokedContext ctx, CancellationToken _) => invokedContexts.Add(ctx))
+            .Returns(() => new ValueTask());
+
+        // Define a simple tool
+        var tool = AIFunctionFactory.Create(() => "tool result", "myTool", "A test tool");
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatOptions = new() { Tools = [tool] },
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        }, services: new ServiceCollection().BuildServiceProvider());
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        Exception? caughtException = null;
+        try
+        {
+            await agent.RunAsync([new(ChatRole.User, "test")], session);
+        }
+        catch (Exception ex)
+        {
+            caughtException = ex;
+        }
+
+        // Diagnostic: check if there was an unexpected exception
+        Assert.Null(caughtException);
+
+        // Assert — the decorator should have been called twice (once per service call in the function invocation loop)
+        Assert.Equal(2, serviceCallCount);
+        Assert.Equal(2, invokedContexts.Count);
+
+        // First invocation should have the user message as request and tool call response
+        Assert.NotNull(invokedContexts[0].ResponseMessages);
+        var firstRequestMessages = invokedContexts[0].RequestMessages.ToList();
+        Assert.Contains(firstRequestMessages, m => m.Text == "test");
+        Assert.Contains(invokedContexts[0].ResponseMessages!, m => m.Contents.OfType<FunctionCallContent>().Any());
+
+        // Second invocation: request messages should NOT include the original user message (already notified).
+        // It should only include messages added since the first call (assistant tool call + tool result).
+        Assert.NotNull(invokedContexts[1].ResponseMessages);
+        var secondRequestMessages = invokedContexts[1].RequestMessages.ToList();
+        Assert.DoesNotContain(secondRequestMessages, m => m.Text == "test");
+        Assert.Contains(invokedContexts[1].ResponseMessages!, m => m.Text == "final response");
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default) with streaming,
+    /// the ChatHistoryProvider receives messages after the stream completes.
+    /// </summary>
+    [Fact]
+    public async Task RunStreamingAsync_PersistsMessagesPerServiceCall_ByDefaultAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetStreamingResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>()))
+            .Returns(CreateAsyncEnumerableAsync(
+                new ChatResponseUpdate(ChatRole.Assistant, "streaming "),
+                new ChatResponseUpdate(ChatRole.Assistant, "response")));
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await foreach (var _ in agent.RunStreamingAsync([new(ChatRole.User, "test")], session))
+        {
+            // Consume stream
+        }
+
+        // Assert — InvokedCoreAsync should be called by the decorator
+        mockChatHistoryProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<ChatHistoryProvider.InvokedContext>(x =>
+                    x.RequestMessages.Any(m => m.Text == "test") &&
+                    x.ResponseMessages != null),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default),
+    /// AIContextProviders are also notified of new messages after a successful call.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_NotifiesAIContextProviders_ByDefaultAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")]));
+
+        Mock<AIContextProvider> mockContextProvider = new(null, null, null);
+        mockContextProvider.SetupGet(p => p.StateKeys).Returns(["TestAIContextProvider"]);
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask<AIContext>>("InvokingCoreAsync", ItExpr.IsAny<AIContextProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask<AIContext>(new AIContext()));
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<AIContextProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            AIContextProviders = [mockContextProvider.Object],
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert — InvokedCoreAsync should be called by the decorator for the AIContextProvider
+        mockContextProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<AIContextProvider.InvokedContext>(x =>
+                    x.ResponseMessages != null &&
+                    x.ResponseMessages.Any(m => m.Text == "response")),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default) and the service fails,
+    /// AIContextProviders are notified of the failure.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_NotifiesAIContextProvidersOfFailure_ByDefaultAsync()
+    {
+        // Arrange
+        var expectedException = new InvalidOperationException("Service failed");
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ThrowsAsync(expectedException);
+
+        Mock<AIContextProvider> mockContextProvider = new(null, null, null);
+        mockContextProvider.SetupGet(p => p.StateKeys).Returns(["TestAIContextProvider"]);
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask<AIContext>>("InvokingCoreAsync", ItExpr.IsAny<AIContextProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask<AIContext>(new AIContext()));
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<AIContextProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            AIContextProviders = [mockContextProvider.Object],
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await Assert.ThrowsAsync<InvalidOperationException>(() => agent.RunAsync([new(ChatRole.User, "test")], session));
+
+        // Assert — the decorator should have notified the AIContextProvider of the failure
+        mockContextProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<AIContextProvider.InvokedContext>(x =>
+                    x.InvokeException != null &&
+                    x.InvokeException.Message == "Service failed"),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is active (default),
+    /// both ChatHistoryProvider and AIContextProviders are notified together.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_NotifiesBothProviders_ByDefaultAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")]));
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask());
+
+        Mock<AIContextProvider> mockContextProvider = new(null, null, null);
+        mockContextProvider.SetupGet(p => p.StateKeys).Returns(["TestAIContextProvider"]);
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask<AIContext>>("InvokingCoreAsync", ItExpr.IsAny<AIContextProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask<AIContext>(new AIContext()));
+        mockContextProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<AIContextProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            AIContextProviders = [mockContextProvider.Object],
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert — both providers should have been notified
+        mockChatHistoryProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<ChatHistoryProvider.InvokedContext>(x =>
+                    x.ResponseMessages != null &&
+                    x.ResponseMessages.Any(m => m.Text == "response")),
+                ItExpr.IsAny<CancellationToken>());
+
+        mockContextProvider
+            .Protected()
+            .Verify<ValueTask>("InvokedCoreAsync", Times.Once(),
+                ItExpr.Is<AIContextProvider.InvokedContext>(x =>
+                    x.ResponseMessages != null &&
+                    x.ResponseMessages.Any(m => m.Text == "response")),
+                ItExpr.IsAny<CancellationToken>());
+    }
+
+    /// <summary>
+    /// Verifies that during a FIC loop, response messages from the first call are not
+    /// re-notified as request messages on the second call.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_DoesNotReNotifyResponseMessagesAsRequestMessages_DuringFicLoopAsync()
+    {
+        // Arrange
+        int serviceCallCount = 0;
+        var assistantToolCallMessage = new ChatMessage(ChatRole.Assistant, [new FunctionCallContent("call1", "myTool", new Dictionary<string, object?>())]);
+
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>()))
+            .Returns(() =>
+            {
+                serviceCallCount++;
+                if (serviceCallCount == 1)
+                {
+                    return Task.FromResult(new ChatResponse([assistantToolCallMessage]));
+                }
+
+                return Task.FromResult(new ChatResponse([new(ChatRole.Assistant, "final response")]));
+            });
+
+        var invokedContexts = new List<ChatHistoryProvider.InvokedContext>();
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Callback((ChatHistoryProvider.InvokedContext ctx, CancellationToken _) => invokedContexts.Add(ctx))
+            .Returns(() => new ValueTask());
+
+        var tool = AIFunctionFactory.Create(() => "tool result", "myTool", "A test tool");
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatOptions = new() { Tools = [tool] },
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        }, services: new ServiceCollection().BuildServiceProvider());
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert
+        Assert.Equal(2, invokedContexts.Count);
+
+        // The assistant tool call message was a response in call 1
+        Assert.Contains(invokedContexts[0].ResponseMessages!, m => ReferenceEquals(m, assistantToolCallMessage));
+
+        // It should NOT appear as a request in call 2 (it was already notified as a response)
+        var secondRequestMessages = invokedContexts[1].RequestMessages.ToList();
+        Assert.DoesNotContain(secondRequestMessages, m => ReferenceEquals(m, assistantToolCallMessage));
+    }
+
+    /// <summary>
+    /// Verifies that when a failure occurs on the second call in a FIC loop,
+    /// only new request messages (not previously notified) are sent in the failure notification.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_DeduplicatesRequestMessages_OnFailureDuringFicLoopAsync()
+    {
+        // Arrange
+        int serviceCallCount = 0;
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>()))
+            .Returns(() =>
+            {
+                serviceCallCount++;
+                if (serviceCallCount == 1)
+                {
+                    return Task.FromResult(new ChatResponse([new(ChatRole.Assistant, [new FunctionCallContent("call1", "myTool", new Dictionary<string, object?>())])]));
+                }
+
+                throw new InvalidOperationException("Service failure on second call");
+            });
+
+        var invokedContexts = new List<ChatHistoryProvider.InvokedContext>();
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Callback((ChatHistoryProvider.InvokedContext ctx, CancellationToken _) => invokedContexts.Add(ctx))
+            .Returns(() => new ValueTask());
+
+        var tool = AIFunctionFactory.Create(() => "tool result", "myTool", "A test tool");
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatOptions = new() { Tools = [tool] },
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        }, services: new ServiceCollection().BuildServiceProvider());
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await Assert.ThrowsAsync<InvalidOperationException>(() =>
+            agent.RunAsync([new(ChatRole.User, "test")], session));
+
+        // Assert — should have 2 notifications: success on call 1, failure on call 2
+        Assert.Equal(2, invokedContexts.Count);
+
+        // First notification: success, has user message as request
+        Assert.Null(invokedContexts[0].InvokeException);
+        Assert.Contains(invokedContexts[0].RequestMessages, m => m.Text == "test");
+
+        // Second notification: failure, should NOT include the user message (already notified)
+        Assert.NotNull(invokedContexts[1].InvokeException);
+        var failureRequestMessages = invokedContexts[1].RequestMessages.ToList();
+        Assert.DoesNotContain(failureRequestMessages, m => m.Text == "test");
+    }
+
+    /// <summary>
+    /// Verifies that after a successful run with per-service-call persistence, the notified
+    /// messages are stamped with the persisted marker so they are not re-notified.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_MarksNotifiedMessages_WithPersistedMarkerAsync()
+    {
+        // Arrange
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>())).ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")]));
+
+        Mock<ChatHistoryProvider> mockChatHistoryProvider = new(null, null, null);
+        mockChatHistoryProvider.SetupGet(p => p.StateKeys).Returns(["TestChatHistoryProvider"]);
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask<IEnumerable<ChatMessage>>>("InvokingCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokingContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns((ChatHistoryProvider.InvokingContext ctx, CancellationToken _) =>
+                new ValueTask<IEnumerable<ChatMessage>>(ctx.RequestMessages.ToList()));
+        mockChatHistoryProvider
+            .Protected()
+            .Setup<ValueTask>("InvokedCoreAsync", ItExpr.IsAny<ChatHistoryProvider.InvokedContext>(), ItExpr.IsAny<CancellationToken>())
+            .Returns(() => new ValueTask());
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            ChatHistoryProvider = mockChatHistoryProvider.Object,
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var inputMessage = new ChatMessage(ChatRole.User, "test");
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([inputMessage], session);
+
+        // Assert — input message should be marked as persisted
+        Assert.True(
+            inputMessage.AdditionalProperties?.ContainsKey(ChatHistoryPersistingChatClient.PersistedMarkerKey) == true,
+            "Input message should be marked as persisted after a successful run.");
+    }
+
+    /// <summary>
+    /// Verifies that when per-service-call persistence is enabled and the inner client returns a
+    /// conversation ID, the session's ConversationId is updated after the service call.
+    /// </summary>
+    [Fact]
+    public async Task RunAsync_UpdatesSessionConversationId_WhenPerServiceCallPersistenceEnabledAsync()
+    {
+        // Arrange
+        const string ExpectedConversationId = "conv-123";
+
+        Mock<IChatClient> mockService = new();
+        mockService.Setup(
+            s => s.GetResponseAsync(
+                It.IsAny<IEnumerable<ChatMessage>>(),
+                It.IsAny<ChatOptions>(),
+                It.IsAny<CancellationToken>()))
+            .ReturnsAsync(new ChatResponse([new(ChatRole.Assistant, "response")])
+            {
+                ConversationId = ExpectedConversationId,
+            });
+
+        ChatClientAgent agent = new(mockService.Object, options: new()
+        {
+            PersistChatHistoryAtEndOfRun = false,
+        });
+
+        // Act
+        var session = await agent.CreateSessionAsync() as ChatClientAgentSession;
+        await agent.RunAsync([new(ChatRole.User, "test")], session);
+
+        // Assert — session should have the conversation ID returned by the inner client
+        Assert.Equal(ExpectedConversationId, session!.ConversationId);
+    }
+
+    private static async IAsyncEnumerable<ChatResponseUpdate> CreateAsyncEnumerableAsync(params ChatResponseUpdate[] updates)
+    {
+        foreach (var update in updates)
+        {
+            yield return update;
+        }
+
+        await Task.CompletedTask;
+    }
+}
@@ -10,20 +10,8 @@ using Microsoft.Extensions.AI;

 namespace Microsoft.Agents.AI.Workflows.UnitTests;

-public class AIAgentHostExecutorTests
+public class AIAgentHostExecutorTests : AIAgentHostingExecutorTestsBase
 {
-    private const string TestAgentId = nameof(TestAgentId);
-    private const string TestAgentName = nameof(TestAgentName);
-
-    private static readonly string[] s_messageStrings = [
-        "",
-        "Hello world!",
-        "Lorem ipsum dolor sit amet, consectetur adipiscing elit.",
-        "Quisque dignissim ante odio, at facilisis orci porta a. Duis mi augue, fringilla eu egestas a, pellentesque sed lacus."
-    ];
-
-    private static List<ChatMessage> TestMessages => TestReplayAgent.ToChatMessages(s_messageStrings);
-
    [Theory]
    [InlineData(null, null)]
    [InlineData(null, true)]
@@ -50,30 +38,7 @@ public class AIAgentHostExecutorTests
        bool expectingEvents = turnSetting ?? executorSetting ?? false;

        AgentResponseUpdateEvent[] updates = testContext.Events.OfType<AgentResponseUpdateEvent>().ToArray();
-        if (expectingEvents)
-        {
-            // The way TestReplayAgent is set up, it will emit one update per non-empty AIContent
-            List<AIContent> expectedUpdateContents = TestMessages.SelectMany(message => message.Contents).ToList();
-
-            updates.Should().HaveCount(expectedUpdateContents.Count);
-            for (int i = 0; i < updates.Length; i++)
-            {
-                AgentResponseUpdateEvent updateEvent = updates[i];
-                AIContent expectedUpdateContent = expectedUpdateContents[i];
-
-                updateEvent.ExecutorId.Should().Be(agent.GetDescriptiveId());
-
-                AgentResponseUpdate update = updateEvent.Update;
-                update.AuthorName.Should().Be(TestAgentName);
-                update.AgentId.Should().Be(TestAgentId);
-                update.Contents.Should().HaveCount(1);
-                update.Contents[0].Should().BeEquivalentTo(expectedUpdateContent);
-            }
-        }
-        else
-        {
-            updates.Should().BeEmpty();
-        }
+        CheckResponseUpdateEventsAgainstTestMessages(updates, expectingEvents, agent.GetDescriptiveId());
    }

    [Theory]
@@ -92,30 +57,7 @@ public class AIAgentHostExecutorTests

        // Assert
        AgentResponseEvent[] updates = testContext.Events.OfType<AgentResponseEvent>().ToArray();
-        if (executorSetting)
-        {
-            updates.Should().HaveCount(1);
-
-            AgentResponseEvent responseEvent = updates[0];
-            responseEvent.ExecutorId.Should().Be(agent.GetDescriptiveId());
-
-            AgentResponse response = responseEvent.Response;
-            response.AgentId.Should().Be(TestAgentId);
-            response.Messages.Should().HaveCount(TestMessages.Count - 1);
-
-            for (int i = 0; i < response.Messages.Count; i++)
-            {
-                ChatMessage responseMessage = response.Messages[i];
-                ChatMessage expectedMessage = TestMessages[i + 1]; // Skip the first empty message
-
-                responseMessage.AuthorName.Should().Be(TestAgentName);
-                responseMessage.Text.Should().Be(expectedMessage.Text);
-            }
-        }
-        else
-        {
-            updates.Should().BeEmpty();
-        }
+        CheckResponseEventsAgainstTestMessages(updates, expectingResponse: executorSetting, agent.GetDescriptiveId());
    }

    private static ChatMessage UserMessage => new(ChatRole.User, "Hello from User!") { AuthorName = "User" };
@@ -0,0 +1,79 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using System.Collections.Generic;
+using System.Linq;
+using FluentAssertions;
+using Microsoft.Extensions.AI;
+
+namespace Microsoft.Agents.AI.Workflows.UnitTests;
+
+public abstract class AIAgentHostingExecutorTestsBase
+{
+    protected const string TestAgentId = nameof(TestAgentId);
+    protected const string TestAgentName = nameof(TestAgentName);
+
+    private static readonly string[] s_messageStrings = [
+        "",
+        "Hello world!",
+        "Lorem ipsum dolor sit amet, consectetur adipiscing elit.",
+        "Quisque dignissim ante odio, at facilisis orci porta a. Duis mi augue, fringilla eu egestas a, pellentesque sed lacus."
+    ];
+
+    protected static List<ChatMessage> TestMessages => TestReplayAgent.ToChatMessages(s_messageStrings);
+
+    protected static void CheckResponseUpdateEventsAgainstTestMessages(AgentResponseUpdateEvent[] updates, bool expectingEvents, string expectedExecutorId)
+    {
+        if (expectingEvents)
+        {
+            // The way TestReplayAgent is set up, it will emit one update per non-empty AIContent
+            List<AIContent> expectedUpdateContents = TestMessages.SelectMany(message => message.Contents).ToList();
+
+            updates.Should().HaveCount(expectedUpdateContents.Count);
+            for (int i = 0; i < updates.Length; i++)
+            {
+                AgentResponseUpdateEvent updateEvent = updates[i];
+                AIContent expectedUpdateContent = expectedUpdateContents[i];
+
+                updateEvent.ExecutorId.Should().Be(expectedExecutorId);
+
+                AgentResponseUpdate update = updateEvent.Update;
+                update.AuthorName.Should().Be(TestAgentName);
+                update.AgentId.Should().Be(TestAgentId);
+                update.Contents.Should().HaveCount(1);
+                update.Contents[0].Should().BeEquivalentTo(expectedUpdateContent);
+            }
+        }
+        else
+        {
+            updates.Should().BeEmpty();
+        }
+    }
+
+    protected static void CheckResponseEventsAgainstTestMessages(AgentResponseEvent[] updates, bool expectingResponse, string expectedExecutorId)
+    {
+        if (expectingResponse)
+        {
+            updates.Should().HaveCount(1);
+
+            AgentResponseEvent responseEvent = updates[0];
+            responseEvent.ExecutorId.Should().Be(expectedExecutorId);
+
+            AgentResponse response = responseEvent.Response;
+            response.AgentId.Should().Be(TestAgentId);
+            response.Messages.Should().HaveCount(TestMessages.Count - 1);
+
+            for (int i = 0; i < response.Messages.Count; i++)
+            {
+                ChatMessage responseMessage = response.Messages[i];
+                ChatMessage expectedMessage = TestMessages[i + 1]; // Skip the first empty message
+
+                responseMessage.AuthorName.Should().Be(TestAgentName);
+                responseMessage.Text.Should().Be(expectedMessage.Text);
+            }
+        }
+        else
+        {
+            updates.Should().BeEmpty();
+        }
+    }
+}
@@ -9,49 +9,173 @@ using Microsoft.Agents.AI.Workflows.Checkpointing;

 namespace Microsoft.Agents.AI.Workflows.UnitTests;

+internal sealed class TempDirectory : IDisposable
+{
+    public DirectoryInfo DirectoryInfo { get; }
+
+    public TempDirectory()
+    {
+        string tempDirPath = Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString());
+        this.DirectoryInfo = Directory.CreateDirectory(tempDirPath);
+    }
+
+    public void Dispose()
+    {
+        this.DisposeInternal();
+        GC.SuppressFinalize(this);
+    }
+
+    private void DisposeInternal()
+    {
+        if (this.DirectoryInfo.Exists)
+        {
+            try
+            {
+                // Best efforts
+                this.DirectoryInfo.Delete(recursive: true);
+            }
+            catch { }
+        }
+    }
+
+    ~TempDirectory()
+    {
+        // Best efforts
+        this.DisposeInternal();
+    }
+
+    public static implicit operator DirectoryInfo(TempDirectory tempDirectory) => tempDirectory.DirectoryInfo;
+
+    public string FullName => this.DirectoryInfo.FullName;
+
+    public bool IsParentOf(FileInfo candidate)
+    {
+        if (candidate.Directory is null)
+        {
+            return false;
+        }
+
+        if (candidate.Directory.FullName == this.DirectoryInfo.FullName)
+        {
+            return true;
+        }
+
+        return this.IsParentOf(candidate.Directory);
+    }
+
+    public bool IsParentOf(DirectoryInfo candidate)
+    {
+        while (candidate.Parent is not null)
+        {
+            if (candidate.Parent.FullName == this.DirectoryInfo.FullName)
+            {
+                return true;
+            }
+
+            candidate = candidate.Parent;
+        }
+
+        return false;
+    }
+}
 public sealed class FileSystemJsonCheckpointStoreTests
 {
+    public static JsonElement TestData => JsonSerializer.SerializeToElement(new { test = "data" });
+
    [Fact]
    public async Task CreateCheckpointAsync_ShouldPersistIndexToDiskBeforeDisposeAsync()
    {
        // Arrange
-        DirectoryInfo tempDir = new(Path.Combine(Path.GetTempPath(), Guid.NewGuid().ToString()));
-        FileSystemJsonCheckpointStore? store = null;
+        using TempDirectory tempDirectory = new();
+        using FileSystemJsonCheckpointStore? store = new(tempDirectory);

-        try
+        string runId = Guid.NewGuid().ToString("N");
+
+        // Act
+        CheckpointInfo checkpoint = await store.CreateCheckpointAsync(runId, TestData);
+
+        // Assert - Check the file size before disposing to verify data was flushed to disk
+        // The index.jsonl file is held exclusively by the store, so we check via FileInfo
+        string indexPath = Path.Combine(tempDirectory.FullName, "index.jsonl");
+        FileInfo indexFile = new(indexPath);
+        indexFile.Refresh();
+        long fileSizeBeforeDispose = indexFile.Length;
+
+        // Data should already be on disk (file size > 0) before we dispose
+        fileSizeBeforeDispose.Should().BeGreaterThan(0, "index.jsonl should be flushed to disk after CreateCheckpointAsync");
+
+        // Dispose to release file lock before final verification
+        store.Dispose();
+
+        string[] lines = File.ReadAllLines(indexPath);
+        lines.Should().HaveCount(1);
+        lines[0].Should().Contain(checkpoint.CheckpointId);
+    }
+
+    private async ValueTask Run_EscapeRootFolderTestAsync(string escapingPath)
+    {
+        // Arrange
+        using TempDirectory tempDirectory = new();
+        using FileSystemJsonCheckpointStore store = new(tempDirectory);
+
+        string naivePath = Path.Combine(tempDirectory.DirectoryInfo.FullName, escapingPath);
+
+        // Check that the naive path is actually outside the temp directory to validate the test is meaningful
+        FileInfo naiveCheckpointFile = new(naivePath);
+        tempDirectory.IsParentOf(naiveCheckpointFile).Should().BeFalse("The naive path should be outside the root folder to validate that escaping is necessary.");
+
+        // Act
+        CheckpointInfo checkpointInfo = await store.CreateCheckpointAsync(escapingPath, TestData);
+
+        // Assert
+        string naivePathWithCheckpointId = Path.Combine(tempDirectory.DirectoryInfo.FullName, $"{escapingPath}_{checkpointInfo.CheckpointId}.json");
+        new FileInfo(naivePathWithCheckpointId).Exists.Should().BeFalse("The naive path should not be used to save a checkpoint file.");
+
+        string actualFileName = store.GetFileNameForCheckpoint(escapingPath, checkpointInfo);
+        string actualFilePath = Path.Combine(tempDirectory.DirectoryInfo.FullName, actualFileName);
+        FileInfo actualFile = new(actualFilePath);
+
+        tempDirectory.IsParentOf(actualFile).Should().BeTrue("The actual checkpoint should be saved inside the root folder.");
+        actualFile.Exists.Should().BeTrue("The actual path should be used to save a checkpoint file.");
+    }
+
+    [Fact]
+    public async Task CreateCheckpointAsync_ShouldNotEscapeRootFolderAsync()
+    {
+        // The SessionId is used as part of the file name, but if it contains path characters such as /.. it can escape the root folder.
+        // Testing that such characters are escaped properly to prevent directory traversal attacks, etc.
+
+        await this.Run_EscapeRootFolderTestAsync("../valid_suffix");
+
+#if !NETFRAMEWORK
+        if (OperatingSystem.IsWindows())
        {
-            store = new(tempDir);
-            string runId = Guid.NewGuid().ToString("N");
-            JsonElement testData = JsonSerializer.SerializeToElement(new { test = "data" });
-
-            // Act
-            CheckpointInfo checkpoint = await store.CreateCheckpointAsync(runId, testData);
-
-            // Assert - Check the file size before disposing to verify data was flushed to disk
-            // The index.jsonl file is held exclusively by the store, so we check via FileInfo
-            string indexPath = Path.Combine(tempDir.FullName, "index.jsonl");
-            FileInfo indexFile = new(indexPath);
-            indexFile.Refresh();
-            long fileSizeBeforeDispose = indexFile.Length;
-
-            // Data should already be on disk (file size > 0) before we dispose
-            fileSizeBeforeDispose.Should().BeGreaterThan(0, "index.jsonl should be flushed to disk after CreateCheckpointAsync");
-
-            // Dispose to release file lock before final verification
-            store.Dispose();
-            store = null;
-
-            string[] lines = File.ReadAllLines(indexPath);
-            lines.Should().HaveCount(1);
-            lines[0].Should().Contain(checkpoint.CheckpointId);
-        }
-        finally
-        {
-            store?.Dispose();
-            if (tempDir.Exists)
-            {
-                tempDir.Delete(recursive: true);
-            }
+            // Windows allows both \ and / as path separators, so we test both
+            await this.Run_EscapeRootFolderTestAsync("..\\valid_suffix");
        }
+#else
+        // .NET Framework is always on Windows
+        await this.Run_EscapeRootFolderTestAsync("..\\valid_suffix");
+#endif
+    }
+
+    private const string InvalidPathCharsWin32 = "\\/:*?\"<>|";
+    private const string InvalidPathCharsUnix = "/";
+    private const string InvalidPathCharsMacOS = "/:";
+
+    [Theory]
+    [InlineData(InvalidPathCharsWin32)]
+    [InlineData(InvalidPathCharsUnix)]
+    [InlineData(InvalidPathCharsMacOS)]
+    public async Task CreateCheckpointAsync_EscapesInvalidCharsAsync(string invalidChars)
+    {
+        // Arrange
+        using TempDirectory tempDirectory = new();
+        using FileSystemJsonCheckpointStore store = new(tempDirectory);
+
+        string runId = $"prefix_{invalidChars}_suffix";
+
+        Func<Task> createCheckpointAction = async () => await store.CreateCheckpointAsync(runId, TestData);
+        await createCheckpointAction.Should().NotThrowAsync();
    }
 }
@@ -0,0 +1,71 @@
+// Copyright (c) Microsoft. All rights reserved.
+
+using System.Linq;
+using System.Threading.Tasks;
+using Microsoft.Agents.AI.Workflows.Specialized;
+
+namespace Microsoft.Agents.AI.Workflows.UnitTests;
+
+public class HandoffAgentExecutorTests : AIAgentHostingExecutorTestsBase
+{
+    [Theory]
+    [InlineData(null, null)]
+    [InlineData(null, true)]
+    [InlineData(null, false)]
+    [InlineData(true, null)]
+    [InlineData(true, true)]
+    [InlineData(true, false)]
+    [InlineData(false, null)]
+    [InlineData(false, true)]
+    [InlineData(false, false)]
+    public async Task Test_HandoffAgentExecutor_EmitsStreamingUpdatesIFFConfiguredAsync(bool? executorSetting, bool? turnSetting)
+    {
+        // Arrange
+        TestRunContext testContext = new();
+        TestReplayAgent agent = new(TestMessages, TestAgentId, TestAgentName);
+
+        HandoffAgentExecutorOptions options = new("",
+                                                  emitAgentResponseEvents: false,
+                                                  emitAgentResponseUpdateEvents: executorSetting,
+                                                  HandoffToolCallFilteringBehavior.None);
+
+        HandoffAgentExecutor executor = new(agent, options);
+        testContext.ConfigureExecutor(executor);
+
+        // Act
+        HandoffState message = new(new(turnSetting), null, []);
+        await executor.HandleAsync(message, testContext.BindWorkflowContext(executor.Id));
+
+        // Assert
+        bool expectingStreamingUpdates = turnSetting ?? executorSetting ?? false;
+
+        AgentResponseUpdateEvent[] updates = testContext.Events.OfType<AgentResponseUpdateEvent>().ToArray();
+        CheckResponseUpdateEventsAgainstTestMessages(updates, expectingStreamingUpdates, agent.GetDescriptiveId());
+    }
+
+    [Theory]
+    [InlineData(true)]
+    [InlineData(false)]
+    public async Task Test_HandoffAgentExecutor_EmitsResponseIFFConfiguredAsync(bool executorSetting)
+    {
+        // Arrange
+        TestRunContext testContext = new();
+        TestReplayAgent agent = new(TestMessages, TestAgentId, TestAgentName);
+
+        HandoffAgentExecutorOptions options = new("",
+                                                  emitAgentResponseEvents: executorSetting,
+                                                  emitAgentResponseUpdateEvents: false,
+                                                  HandoffToolCallFilteringBehavior.None);
+
+        HandoffAgentExecutor executor = new(agent, options);
+        testContext.ConfigureExecutor(executor);
+
+        // Act
+        HandoffState message = new(new(false), null, []);
+        await executor.HandleAsync(message, testContext.BindWorkflowContext(executor.Id));
+
+        // Assert
+        AgentResponseEvent[] updates = testContext.Events.OfType<AgentResponseEvent>().ToArray();
+        CheckResponseEventsAgainstTestMessages(updates, expectingResponse: executorSetting, agent.GetDescriptiveId());
+    }
+}
@@ -673,6 +673,55 @@ public class JsonSerializationTests
        ValidateCheckpoint(retrievedCheckpoint, prototype);
    }

+    [Fact]
+    public void Test_SessionState_JsonRoundtrip_WithPendingRequests()
+    {
+        // Arrange
+        Dictionary<string, ExternalRequest> pendingRequests = new()
+        {
+            ["call-1"] = TestExternalRequest,
+            ["call-2"] = ExternalRequest.Create(TestPort, "Request2", "OtherData"),
+        };
+
+        WorkflowSession.SessionState prototype = new(
+            sessionId: "test-session-123",
+            lastCheckpoint: TestParentCheckpointInfo,
+            pendingRequests: pendingRequests);
+
+        // Act
+        WorkflowSession.SessionState result = RunJsonRoundtrip(prototype);
+
+        // Assert
+        result.SessionId.Should().Be(prototype.SessionId);
+        result.LastCheckpoint.Should().Be(prototype.LastCheckpoint);
+        result.StateBag.Should().NotBeNull();
+        result.PendingRequests.Should().NotBeNull()
+            .And.HaveCount(pendingRequests.Count);
+
+        foreach (string key in pendingRequests.Keys)
+        {
+            result.PendingRequests.Should().ContainKey(key);
+            ValidateExternalRequest(result.PendingRequests![key], pendingRequests[key]);
+        }
+    }
+
+    [Fact]
+    public void Test_SessionState_JsonRoundtrip_WithoutPendingRequests()
+    {
+        // Arrange
+        WorkflowSession.SessionState prototype = new(
+            sessionId: "test-session-456",
+            lastCheckpoint: null);
+
+        // Act
+        WorkflowSession.SessionState result = RunJsonRoundtrip(prototype);
+
+        // Assert
+        result.SessionId.Should().Be(prototype.SessionId);
+        result.LastCheckpoint.Should().BeNull();
+        result.PendingRequests.Should().BeNull();
+    }
+
    /// <summary>
    /// Verifies that the default behavior (without AllowOutOfOrderMetadataProperties) fails
    /// when $type metadata is not the first property, demonstrating the PostgreSQL jsonb issue.
@@ -28,7 +28,185 @@ public sealed class ExpectedException : Exception
    }
 }

-public class WorkflowHostSmokeTests
+/// <summary>
+/// A simple agent that emits a FunctionCallContent or ToolApprovalRequestContent request.
+/// Used to test that RequestInfoEvent handling preserves the original content type.
+/// </summary>
+internal sealed class RequestEmittingAgent : AIAgent
+{
+    private readonly AIContent _requestContent;
+    private readonly bool _completeOnResponse;
+
+    /// <summary>
+    /// Creates a new <see cref="RequestEmittingAgent"/> that emits the given request content.
+    /// </summary>
+    /// <param name="requestContent">The content to emit on each turn.</param>
+    /// <param name="completeOnResponse">
+    /// When <see langword="true"/>, the agent emits a text completion instead of re-emitting
+    /// the request when the incoming messages contain a <see cref="FunctionResultContent"/>
+    /// or <see cref="ToolApprovalResponseContent"/>.  This models realistic agent behaviour
+    /// where the agent processes the tool result and produces a final answer.
+    /// </param>
+    public RequestEmittingAgent(AIContent requestContent, bool completeOnResponse = false)
+    {
+        this._requestContent = requestContent;
+        this._completeOnResponse = completeOnResponse;
+    }
+
+    private sealed class Session : AgentSession
+    {
+        public Session() { }
+    }
+
+    protected override ValueTask<AgentSession> DeserializeSessionCoreAsync(JsonElement serializedState, JsonSerializerOptions? jsonSerializerOptions = null, CancellationToken cancellationToken = default)
+        => new(new Session());
+
+    protected override ValueTask<AgentSession> CreateSessionCoreAsync(CancellationToken cancellationToken = default)
+        => new(new Session());
+
+    protected override ValueTask<JsonElement> SerializeSessionCoreAsync(AgentSession session, JsonSerializerOptions? jsonSerializerOptions = null, CancellationToken cancellationToken = default)
+        => default;
+
+    protected override Task<AgentResponse> RunCoreAsync(IEnumerable<ChatMessage> messages, AgentSession? session = null, AgentRunOptions? options = null, CancellationToken cancellationToken = default)
+        => this.RunStreamingAsync(messages, session, options, cancellationToken).ToAgentResponseAsync(cancellationToken);
+
+    protected override async IAsyncEnumerable<AgentResponseUpdate> RunCoreStreamingAsync(IEnumerable<ChatMessage> messages, AgentSession? session = null, AgentRunOptions? options = null, [EnumeratorCancellation] CancellationToken cancellationToken = default)
+    {
+        if (this._completeOnResponse && messages.Any(m => m.Contents.Any(c =>
+            c is FunctionResultContent || c is ToolApprovalResponseContent)))
+        {
+            yield return new AgentResponseUpdate(ChatRole.Assistant, [new TextContent("Request processed")]);
+        }
+        else
+        {
+            // Emit the request content
+            yield return new AgentResponseUpdate(ChatRole.Assistant, [this._requestContent]);
+        }
+    }
+}
+
+internal sealed class KickoffOnStartExecutor : ChatProtocolExecutor
+{
+    private static readonly ChatProtocolExecutorOptions s_options = new()
+    {
+        AutoSendTurnToken = false,
+    };
+
+    private readonly string _downstreamExecutorId;
+    private readonly string _kickoffInputText;
+    private readonly string _kickoffMessageText;
+    private readonly string _regularResumeText;
+    private readonly string _regularProcessedText;
+
+    public KickoffOnStartExecutor(
+        string id,
+        string downstreamExecutorId,
+        string kickoffInputText,
+        string kickoffMessageText,
+        string regularResumeText,
+        string regularProcessedText)
+        : base(id, s_options)
+    {
+        this._downstreamExecutorId = downstreamExecutorId;
+        this._kickoffInputText = kickoffInputText;
+        this._kickoffMessageText = kickoffMessageText;
+        this._regularResumeText = regularResumeText;
+        this._regularProcessedText = regularProcessedText;
+    }
+
+    protected override async ValueTask TakeTurnAsync(List<ChatMessage> messages, IWorkflowContext context, bool? emitEvents, CancellationToken cancellationToken = default)
+    {
+        List<string> textContents =
+        [
+            .. messages
+                .SelectMany(message => message.Contents.OfType<TextContent>())
+                .Select(content => content.Text)
+        ];
+
+        if (textContents.Contains(this._kickoffInputText, StringComparer.Ordinal))
+        {
+            await context.SendMessageAsync(
+                new List<ChatMessage> { new(ChatRole.User, this._kickoffMessageText) },
+                this._downstreamExecutorId,
+                cancellationToken).ConfigureAwait(false);
+            await context.SendMessageAsync(
+                new TurnToken(emitEvents),
+                this._downstreamExecutorId,
+                cancellationToken).ConfigureAwait(false);
+        }
+
+        if (textContents.Contains(this._regularResumeText, StringComparer.Ordinal))
+        {
+            AgentResponseUpdate update = new(ChatRole.Assistant, [new TextContent(this._regularProcessedText)])
+            {
+                CreatedAt = DateTimeOffset.UtcNow,
+                MessageId = Guid.NewGuid().ToString("N"),
+                ResponseId = Guid.NewGuid().ToString("N"),
+                Role = ChatRole.Assistant,
+            };
+
+            await context.AddEventAsync(new AgentResponseUpdateEvent(this.Id, update), cancellationToken).ConfigureAwait(false);
+        }
+    }
+}
+
+/// <summary>
+/// A start executor that always emits a response update on every turn,
+/// useful for verifying that a TurnToken was delivered by the session.
+/// On the first turn (user messages present), it kicks off a downstream executor.
+/// </summary>
+internal sealed class TurnTrackingStartExecutor : ChatProtocolExecutor
+{
+    private static readonly ChatProtocolExecutorOptions s_options = new()
+    {
+        AutoSendTurnToken = false,
+    };
+
+    private readonly string _downstreamExecutorId;
+    private readonly string _activatedMarker;
+    private int _activationCount;
+
+    /// <summary>Gets the number of times this executor has been activated (i.e., <see cref="TakeTurnAsync"/> called).</summary>
+    public int ActivationCount => this._activationCount;
+
+    public TurnTrackingStartExecutor(string id, string downstreamExecutorId, string activatedMarker)
+        : base(id, s_options)
+    {
+        this._downstreamExecutorId = downstreamExecutorId;
+        this._activatedMarker = activatedMarker;
+    }
+
+    protected override async ValueTask TakeTurnAsync(List<ChatMessage> messages, IWorkflowContext context, bool? emitEvents, CancellationToken cancellationToken = default)
+    {
+        Interlocked.Increment(ref this._activationCount);
+
+        // On the first turn, forward user messages and a TurnToken to the downstream executor.
+        if (messages.Any(m => m.Role == ChatRole.User))
+        {
+            await context.SendMessageAsync(
+                messages,
+                this._downstreamExecutorId,
+                cancellationToken).ConfigureAwait(false);
+            await context.SendMessageAsync(
+                new TurnToken(emitEvents),
+                this._downstreamExecutorId,
+                cancellationToken).ConfigureAwait(false);
+        }
+
+        // Always emit a marker to prove this executor was activated.
+        AgentResponseUpdate update = new(ChatRole.Assistant, [new TextContent(this._activatedMarker)])
+        {
+            CreatedAt = DateTimeOffset.UtcNow,
+            MessageId = Guid.NewGuid().ToString("N"),
+            ResponseId = Guid.NewGuid().ToString("N"),
+            Role = ChatRole.Assistant,
+        };
+
+        await context.AddEventAsync(new AgentResponseUpdateEvent(this.Id, update), cancellationToken).ConfigureAwait(false);
+    }
+}
+
+public class WorkflowHostSmokeTests : AIAgentHostingExecutorTestsBase
 {
    private sealed class AlwaysFailsAIAgent(bool failByThrowing) : AIAgent
    {
@@ -112,4 +290,511 @@ public class WorkflowHostSmokeTests

        hadErrorContent.Should().BeTrue();
    }
+
+    /// <summary>
+    /// Tests that when a workflow emits a RequestInfoEvent with FunctionCallContent data,
+    /// the AgentResponseUpdate preserves the original FunctionCallContent type.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_FunctionCallContentPreservedInRequestInfoAsync()
+    {
+        // Arrange
+        const string CallId = "test-call-id";
+        const string FunctionName = "testFunction";
+        FunctionCallContent originalContent = new(CallId, FunctionName);
+        RequestEmittingAgent requestAgent = new(originalContent);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+
+        // Act
+        List<AgentResponseUpdate> updates = await workflow.AsAIAgent("WorkflowAgent")
+                                                           .RunStreamingAsync(new ChatMessage(ChatRole.User, "Hello"))
+                                                           .ToListAsync();
+
+        // Assert
+        AgentResponseUpdate? updateWithFunctionCall = updates.FirstOrDefault(u =>
+            u.RawRepresentation is RequestInfoEvent && u.Contents.Any(c => c is FunctionCallContent));
+
+        updateWithFunctionCall.Should().NotBeNull("a FunctionCallContent should be present in the response updates");
+        FunctionCallContent retrievedContent = updateWithFunctionCall!.Contents
+            .OfType<FunctionCallContent>()
+            .Should().ContainSingle()
+            .Which;
+
+        retrievedContent.CallId.Should().NotBe(CallId);
+        retrievedContent.CallId.Should().EndWith($":{CallId}");
+        retrievedContent.Name.Should().Be(FunctionName);
+    }
+
+    /// <summary>
+    /// Tests that when a workflow emits a RequestInfoEvent with ToolApprovalRequestContent data,
+    /// the AgentResponseUpdate preserves the original ToolApprovalRequestContent type.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_ToolApprovalRequestContentPreservedInRequestInfoAsync()
+    {
+        // Arrange
+        const string RequestId = "test-request-id";
+        McpServerToolCallContent mcpCall = new("call-id", "testToolName", "http://localhost");
+        ToolApprovalRequestContent originalContent = new(RequestId, mcpCall);
+        RequestEmittingAgent requestAgent = new(originalContent);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUserInputRequests = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+
+        // Act
+        List<AgentResponseUpdate> updates = await workflow.AsAIAgent("WorkflowAgent")
+                                                           .RunStreamingAsync(new ChatMessage(ChatRole.User, "Hello"))
+                                                           .ToListAsync();
+
+        // Assert
+        AgentResponseUpdate? updateWithUserInput = updates.FirstOrDefault(u =>
+            u.RawRepresentation is RequestInfoEvent && u.Contents.Any(c => c is ToolApprovalRequestContent));
+
+        updateWithUserInput.Should().NotBeNull("a ToolApprovalRequestContent should be present in the response updates");
+        ToolApprovalRequestContent retrievedContent = updateWithUserInput!.Contents
+            .OfType<ToolApprovalRequestContent>()
+            .Should().ContainSingle()
+            .Which;
+
+        retrievedContent.Should().NotBeNull();
+        retrievedContent.RequestId.Should().NotBe(RequestId);
+        retrievedContent.RequestId.Should().EndWith($":{RequestId}");
+    }
+
+    /// <summary>
+    /// Tests the full roundtrip: workflow emits a request, external caller responds, workflow processes response.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_FunctionCallRoundtrip_ResponseIsProcessedAsync()
+    {
+        // Arrange: Create an agent that emits a FunctionCallContent request
+        const string CallId = "roundtrip-call-id";
+        const string FunctionName = "testFunction";
+        FunctionCallContent requestContent = new(CallId, FunctionName);
+        RequestEmittingAgent requestAgent = new(requestContent, completeOnResponse: true);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        // Act 1: First call - should receive the FunctionCallContent request
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.User, "Start"),
+            session).ToListAsync();
+
+        // Assert 1: We should have received a FunctionCallContent
+        AgentResponseUpdate? updateWithRequest = firstCallUpdates.FirstOrDefault(u =>
+            u.RawRepresentation is RequestInfoEvent && u.Contents.Any(c => c is FunctionCallContent));
+        updateWithRequest.Should().NotBeNull("a FunctionCallContent should be present in the response updates");
+
+        FunctionCallContent receivedRequest = updateWithRequest!.Contents
+            .OfType<FunctionCallContent>()
+            .First();
+        receivedRequest.CallId.Should().EndWith($":{CallId}");
+
+        // Act 2: Send the response back
+        FunctionResultContent responseContent = new(receivedRequest.CallId, "test result");
+        ChatMessage responseMessage = new(ChatRole.Tool, [responseContent]);
+
+        // Act 2: Run the workflow with the response and capture the resulting updates
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(responseMessage, session).ToListAsync();
+
+        // Assert 2: The response should be processed and the original request should no longer be pending.
+        // Concretely, the workflow should not re-emit a FunctionCallContent with the same CallId.
+        secondCallUpdates.Should().NotBeNull("processing the response should produce updates");
+        secondCallUpdates.Should().NotBeEmpty("processing the response should progress the workflow");
+        secondCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Should()
+            .NotContain(c => c.CallId == receivedRequest.CallId, "the external FunctionCallContent request should be cleared after processing the response");
+    }
+
+    /// <summary>
+    /// Tests the full roundtrip for ToolApprovalRequestContent: workflow emits request, external caller responds.
+    /// Verifying inbound ToolApprovalResponseContent conversion.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_ToolApprovalRoundtrip_ResponseIsProcessedAsync()
+    {
+        // Arrange: Create an agent that emits a ToolApprovalRequestContent request
+        const string RequestId = "roundtrip-request-id";
+        McpServerToolCallContent mcpCall = new("mcp-call-id", "testMcpTool", "http://localhost");
+        ToolApprovalRequestContent requestContent = new(RequestId, mcpCall);
+        RequestEmittingAgent requestAgent = new(requestContent, completeOnResponse: true);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUserInputRequests = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        // Act 1: First call - should receive the ToolApprovalRequestContent request
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.User, "Start"),
+            session).ToListAsync();
+
+        // Assert 1: We should have received a ToolApprovalRequestContent
+        AgentResponseUpdate? updateWithRequest = firstCallUpdates.FirstOrDefault(u =>
+            u.RawRepresentation is RequestInfoEvent && u.Contents.Any(c => c is ToolApprovalRequestContent));
+        updateWithRequest.Should().NotBeNull("a ToolApprovalRequestContent should be present in the response updates");
+
+        ToolApprovalRequestContent receivedRequest = updateWithRequest!.Contents
+            .OfType<ToolApprovalRequestContent>()
+            .First();
+        receivedRequest.RequestId.Should().EndWith($":{RequestId}");
+
+        // Act 2: Send the response back - use CreateResponse to get the right response type
+        ToolApprovalResponseContent responseContent = receivedRequest.CreateResponse(approved: true);
+        ChatMessage responseMessage = new(ChatRole.User, [responseContent]);
+
+        // Act 2: Run the workflow again with the response and capture the updates
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(responseMessage, session).ToListAsync();
+
+        // Assert 2: The response should be applied so that the original request is no longer pending
+        secondCallUpdates.Should().NotBeEmpty("handling the user input response should produce follow-up updates");
+        bool requestStillPresent = secondCallUpdates.Any(u =>
+            u.RawRepresentation is RequestInfoEvent
+            && u.Contents.OfType<ToolApprovalRequestContent>().Any(r => r.RequestId == receivedRequest.RequestId));
+        requestStillPresent.Should().BeFalse("the original ToolApprovalRequestContent should not be re-emitted after its response is processed");
+    }
+
+    /// <summary>
+    /// Tests the mixed-message scenario: resume contains both an external response
+    /// (FunctionResultContent matching a pending request) and regular non-response content
+    /// in the same message.
+    /// Verifies that regular content is still processed and that no duplicate
+    /// pending-request errors, redundant FunctionCallContent re-emissions,
+    /// or workflow errors occur.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_MixedResponseAndRegularMessage_BothProcessedAsync()
+    {
+        // Arrange: Create an agent that emits a FunctionCallContent request
+        const string CallId = "mixed-call-id";
+        const string FunctionName = "mixedTestFunction";
+        FunctionCallContent requestContent = new(CallId, FunctionName);
+        RequestEmittingAgent requestAgent = new(requestContent, completeOnResponse: true);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        // Act 1: First call - should receive the FunctionCallContent request
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.User, "Start"),
+            session).ToListAsync();
+
+        // Assert 1: We should have received a FunctionCallContent
+        AgentResponseUpdate requestUpdate = firstCallUpdates.First(u =>
+            u.RawRepresentation is RequestInfoEvent && u.Contents.Any(c => c is FunctionCallContent));
+        FunctionCallContent emittedRequest = requestUpdate.Contents.OfType<FunctionCallContent>().Single();
+
+        firstCallUpdates.Should().Contain(u => u.Contents.Any(c => c is FunctionCallContent),
+            "the first call should emit a FunctionCallContent request");
+
+        // Act 2: Send a mixed message containing both the function result AND regular non-response content
+        FunctionResultContent responseContent = new(emittedRequest.CallId, "tool output");
+        ChatMessage mixedMessage = new(ChatRole.Tool, [responseContent, new TextContent("additional context")]);
+
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(mixedMessage, session).ToListAsync();
+
+        // Assert 2: The workflow should have processed both parts without errors
+        secondCallUpdates.Should().NotBeEmpty("the mixed message should produce follow-up updates");
+        secondCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Should()
+            .NotContain(c => c.CallId == emittedRequest.CallId, "the external FunctionCallContent should be cleared after the response is processed");
+        secondCallUpdates
+            .SelectMany(u => u.Contents.OfType<ErrorContent>())
+            .Should()
+            .BeEmpty("no workflow errors should occur when processing a mixed response-and-regular message");
+    }
+
+    [Fact]
+    public async Task Test_AsAgent_ResponseThenRegularAcrossMessages_NoDuplicateFunctionCallAsync()
+    {
+        const string CallId = "mixed-separate-call-id";
+        const string FunctionName = "mixedSeparateTestFunction";
+
+        RequestEmittingAgent requestAgent = new(new FunctionCallContent(CallId, FunctionName), completeOnResponse: true);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(new ChatMessage(ChatRole.User, "Start"), session).ToListAsync();
+        FunctionCallContent emittedRequest = firstCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Single();
+
+        ChatMessage[] resumeMessages =
+        [
+            new(ChatRole.Tool, [new FunctionResultContent(emittedRequest.CallId, "tool output")]),
+            new(ChatRole.Tool, [new TextContent("extra context in separate message")])
+        ];
+
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(resumeMessages, session).ToListAsync();
+
+        secondCallUpdates.Should().NotBeEmpty();
+        secondCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Should()
+            .NotContain(c => c.CallId == emittedRequest.CallId, "response+regular content split across messages should not re-emit the handled external request");
+        secondCallUpdates
+            .SelectMany(u => u.Contents.OfType<ErrorContent>())
+            .Should()
+            .BeEmpty();
+    }
+
+    [Fact]
+    public async Task Test_AsAgent_MatchingResponse_DoesNotCauseExtraTurnAsync()
+    {
+        const string CallId = "matching-response-call-id";
+        const string FunctionName = "matchingResponseFunction";
+
+        RequestEmittingAgent requestAgent = new(new FunctionCallContent(CallId, FunctionName), completeOnResponse: false);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(new ChatMessage(ChatRole.User, "Start"), session).ToListAsync();
+        FunctionCallContent emittedRequest = firstCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Single();
+
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.Tool, [new FunctionResultContent(emittedRequest.CallId, "tool output")]),
+            session).ToListAsync();
+
+        int functionCallCount = secondCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Count(c => c.CallId == emittedRequest.CallId);
+
+        functionCallCount.Should().Be(1, "a matching external response should not trigger an extra TurnToken-driven turn");
+    }
+
+    [Fact]
+    public async Task Test_AsAgent_MixedResponseAndRegularMessage_CrossExecutorStartExecutorIsReawakenedAsync()
+    {
+        const string StartExecutorId = "start-executor";
+        const string KickoffInputText = "Start";
+        const string KickoffMessageText = "kickoff downstream";
+        const string ResumeRegularText = "resume regular";
+        const string ResumeProcessedText = "regular message processed";
+        const string CallId = "cross-executor-call-id";
+        const string FunctionName = "crossExecutorFunction";
+
+        RequestEmittingAgent requestAgent = new(new FunctionCallContent(CallId, FunctionName), completeOnResponse: true);
+        ExecutorBinding requestBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+
+        KickoffOnStartExecutor startExecutor = new(
+            StartExecutorId,
+            requestBinding.Id,
+            KickoffInputText,
+            KickoffMessageText,
+            ResumeRegularText,
+            ResumeProcessedText);
+        ExecutorBinding startBinding = startExecutor.BindExecutor();
+
+        Workflow workflow = new WorkflowBuilder(startBinding)
+            .AddEdge<List<ChatMessage>>(startBinding, requestBinding, messages =>
+                messages?.Any(message => message.Contents.OfType<TextContent>().Any(content => content.Text == KickoffMessageText)) == true)
+            .AddEdge<TurnToken>(startBinding, requestBinding, _ => true)
+            .Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.User, KickoffInputText),
+            session).ToListAsync();
+        FunctionCallContent emittedRequest = firstCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Single();
+
+        ChatMessage[] resumeMessages =
+        [
+            new(ChatRole.Tool, [new FunctionResultContent(emittedRequest.CallId, "tool output")]),
+            new(ChatRole.User, ResumeRegularText)
+        ];
+
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(resumeMessages, session).ToListAsync();
+        List<string> textContents = [.. secondCallUpdates.SelectMany(update => update.Contents.OfType<TextContent>()).Select(content => content.Text)];
+
+        textContents.Should().Contain(ResumeProcessedText, "the start executor should receive an explicit TurnToken when the matched response wakes a different executor");
+        textContents.Should().Contain("Request processed", "the matched external response should still be delivered to the downstream request owner");
+        secondCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Should()
+            .NotContain(c => c.CallId == emittedRequest.CallId, "the handled external request should not be re-emitted while waking the start executor");
+        secondCallUpdates.SelectMany(u => u.Contents.OfType<ErrorContent>()).Should().BeEmpty();
+    }
+
+    [Fact]
+    public async Task Test_AsAgent_UnmatchedResponse_TriggersTurnAndKeepsProgressingAsync()
+    {
+        const string CallId = "unmatched-response-call-id";
+        const string FunctionName = "unmatchedResponseFunction";
+
+        RequestEmittingAgent requestAgent = new(new FunctionCallContent(CallId, FunctionName), completeOnResponse: false);
+        ExecutorBinding agentBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+        Workflow workflow = new WorkflowBuilder(agentBinding).Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(new ChatMessage(ChatRole.User, "Start"), session).ToListAsync();
+        firstCallUpdates.Should().Contain(u => u.Contents.Any(c => c is FunctionCallContent));
+
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.Tool, [new FunctionResultContent("different-call-id", "tool output")]),
+            session).ToListAsync();
+
+        int functionCallCount = secondCallUpdates
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Count(c => c.CallId == CallId);
+
+        functionCallCount.Should().Be(1, "an unmatched response should be treated as regular input and still drive a TurnToken continuation without workflow errors");
+        secondCallUpdates.SelectMany(u => u.Contents.OfType<ErrorContent>()).Should().BeEmpty();
+    }
+
+    /// <summary>
+    /// Tests that when a resume contains only an external response directed at a non-start executor
+    /// (no regular messages), the start executor still receives a TurnToken and is activated.
+    /// This is a regression test for the case where the TurnToken was previously skipped because
+    /// <c>HasRegularMessages</c> was <see langword="false"/>, leaving the start executor dormant.
+    /// </summary>
+    [Fact]
+    public async Task Test_AsAgent_ResponseOnlyToNonStartExecutor_StartExecutorIsStillActivatedAsync()
+    {
+        // Arrange
+        const string StartExecutorId = "start-executor";
+        const string ActivatedMarker = "start-executor-activated";
+        const string CallId = "response-only-call-id";
+        const string FunctionName = "responseOnlyFunction";
+
+        RequestEmittingAgent requestAgent = new(new FunctionCallContent(CallId, FunctionName), completeOnResponse: true);
+        ExecutorBinding requestBinding = requestAgent.BindAsExecutor(
+            new AIAgentHostOptions { InterceptUnterminatedFunctionCalls = false, EmitAgentUpdateEvents = true });
+
+        TurnTrackingStartExecutor startExecutor = new(StartExecutorId, requestBinding.Id, ActivatedMarker);
+        ExecutorBinding startBinding = startExecutor.BindExecutor();
+
+        Workflow workflow = new WorkflowBuilder(startBinding)
+            .AddEdge<List<ChatMessage>>(startBinding, requestBinding, messages =>
+                messages?.Any(m => m.Contents.OfType<TextContent>().Any()) == true)
+            .AddEdge<TurnToken>(startBinding, requestBinding, _ => true)
+            .Build();
+        AIAgent agent = workflow.AsAIAgent("WorkflowAgent");
+
+        // Act 1: First call triggers the downstream FunctionCallContent request
+        AgentSession session = await agent.CreateSessionAsync();
+        List<AgentResponseUpdate> firstCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.User, "Start"),
+            session).ToListAsync();
+
+        FunctionCallContent emittedRequest = firstCallUpdates
+            .Where(u => u.RawRepresentation is RequestInfoEvent)
+            .SelectMany(u => u.Contents.OfType<FunctionCallContent>())
+            .Single();
+
+        // Act 2: Resume with ONLY the external response (no regular messages)
+        List<AgentResponseUpdate> secondCallUpdates = await agent.RunStreamingAsync(
+            new ChatMessage(ChatRole.Tool, [new FunctionResultContent(emittedRequest.CallId, "tool output")]),
+            session).ToListAsync();
+
+        // Assert: Both the downstream and start executor should have been activated
+        List<string> textContents = [.. secondCallUpdates
+            .SelectMany(u => u.Contents.OfType<TextContent>())
+            .Select(c => c.Text)];
+
+        textContents.Should().Contain("Request processed",
+            "the downstream executor should process the external response");
+        textContents.Should().Contain(ActivatedMarker,
+            "the start executor should receive a TurnToken and be activated even when resume contains only an external response");
+        secondCallUpdates
+            .SelectMany(u => u.Contents.OfType<ErrorContent>())
+            .Should()
+            .BeEmpty();
+    }
+
+    private async Task Run_AsAgent_OutgoingMessagesInHistoryAsync(Workflow workflow, bool runAsync)
+    {
+        // Arrange
+        AIAgent workflowAgent = workflow.AsAIAgent();
+
+        // Act
+        AgentSession session = await workflowAgent.CreateSessionAsync();
+        AgentResponse response;
+        if (runAsync)
+        {
+            List<AgentResponseUpdate> updates = [];
+            await foreach (AgentResponseUpdate update in workflowAgent.RunStreamingAsync(session))
+            {
+                // Skip WorkflowEvent updates, which do not get persisted in ChatHistory; we cannot skip
+                // them after because of a deleterious interaction with .ToAgentResponse() due to the
+                // empty initial message (which is created without a MessageId). When running through the
+                // message merger, it does the right thing internally.
+                if (!string.IsNullOrEmpty(update.Text))
+                {
+                    updates.Add(update);
+                }
+            }
+
+            response = updates.ToAgentResponse();
+        }
+        else
+        {
+            response = await workflowAgent.RunAsync(session);
+        }
+
+        // Assert
+        WorkflowSession workflowSession = session.Should().BeOfType<WorkflowSession>().Subject;
+
+        ChatMessage[] responseMessages = response.Messages.Where(message => message.Contents.Any())
+                                                          .ToArray();
+
+        ChatMessage[] sessionMessages = workflowSession.ChatHistoryProvider.GetAllMessages(workflowSession)
+                                                                           .ToArray();
+
+        // Since we never sent an incoming message, the expectation is that there should be nothing in the session
+        // except the response
+        responseMessages.Should().BeEquivalentTo(sessionMessages, options => options.WithStrictOrdering());
+    }
+
+    [Theory]
+    [InlineData(false)]
+    [InlineData(true)]
+    public Task Test_SingleAgent_AsAgent_OutgoingMessagesInHistoryAsync(bool runAsync)
+    {
+        // Arrange
+        TestReplayAgent agent = new(TestMessages, TestAgentId, TestAgentName);
+        Workflow singleAgentWorkflow = new WorkflowBuilder(agent).Build();
+        return this.Run_AsAgent_OutgoingMessagesInHistoryAsync(singleAgentWorkflow, runAsync);
+    }
+
+    [Theory]
+    [InlineData(false)]
+    [InlineData(true)]
+    public Task Test_Handoffs_AsAgent_OutgoingMessagesInHistoryAsync(bool runAsync)
+    {
+        // Arrange
+        TestReplayAgent agent = new(TestMessages, TestAgentId, TestAgentName);
+        Workflow handoffWorkflow = new HandoffsWorkflowBuilder(agent).Build();
+        return this.Run_AsAgent_OutgoingMessagesInHistoryAsync(handoffWorkflow, runAsync);
+    }
 }
@@ -24,14 +24,14 @@ If you only need specific integrations, you can install at a more granular level
 # also includes workflows and orchestrations
 pip install agent-framework-core --pre

-# Core + Azure AI integration
-pip install agent-framework-azure-ai --pre
+# Core + Azure AI Foundry integration
+pip install agent-framework-foundry --pre

 # Core + Microsoft Copilot Studio integration
 pip install agent-framework-copilotstudio --pre

-# Core + both Microsoft Copilot Studio and Azure AI integration
-pip install agent-framework-microsoft agent-framework-azure-ai --pre
+# Core + both Microsoft Copilot Studio and Azure AI Foundry integration
+pip install agent-framework-microsoft agent-framework-foundry --pre
 ```

 This selective approach is useful when you know which integrations you need, and it is the recommended way to set up lightweight environments.
@@ -53,8 +53,8 @@ AZURE_OPENAI_API_KEY=...
 AZURE_OPENAI_ENDPOINT=...
 AZURE_OPENAI_CHAT_DEPLOYMENT_NAME=...
 ...
-AZURE_AI_PROJECT_ENDPOINT=...
-AZURE_AI_MODEL_DEPLOYMENT_NAME=...
+FOUNDRY_PROJECT_ENDPOINT=...
+FOUNDRY_MODEL=...
 ```

 You can also override environment variables by explicitly passing configuration parameters to the chat client constructor:
@@ -15,7 +15,7 @@ The Azure AI Search integration provides context providers for RAG (Retrieval Au

 ### Basic Usage Example

-See the [Azure AI Search context provider examples](../../samples/02-agents/providers/azure_ai/) which demonstrate:
+See the [Azure AI Search context provider examples](../../samples/02-agents/context_providers/azure_ai_search/) which demonstrate:

 - Semantic search with hybrid (vector + keyword) queries
 - Agentic mode with Knowledge Bases for complex multi-hop reasoning
@@ -16,8 +16,7 @@ from typing import TYPE_CHECKING, Any, ClassVar, Literal, TypedDict
 from agent_framework import AGENT_FRAMEWORK_USER_AGENT, Annotation, Content, Message, SupportsGetEmbeddings
 from agent_framework._sessions import AgentSession, BaseContextProvider, SessionContext
 from agent_framework._settings import SecretString, load_settings
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
-from azure.core.credentials import AzureKeyCredential
+from azure.core.credentials import AzureKeyCredential, TokenCredential
 from azure.core.credentials_async import AsyncTokenCredential
 from azure.core.exceptions import ResourceNotFoundError
 from azure.search.documents.aio import SearchClient
@@ -111,6 +110,8 @@ try:
 except ImportError:
    _agentic_retrieval_available = False

+AzureCredentialTypes = TokenCredential | AsyncTokenCredential
+
 logger = logging.getLogger("agent_framework.azure_ai_search")

 _DEFAULT_AGENTIC_MESSAGE_HISTORY_COUNT = 10
@@ -2,23 +2,35 @@

 import importlib.metadata

-from ._agent_provider import AzureAIAgentsProvider
-from ._chat_client import AzureAIAgentClient, AzureAIAgentOptions
-from ._client import AzureAIClient, AzureAIProjectAgentOptions, RawAzureAIClient
+from ._agent_provider import AzureAIAgentsProvider  # pyright: ignore[reportDeprecated]
+from ._chat_client import AzureAIAgentClient, AzureAIAgentOptions  # pyright: ignore[reportDeprecated]
+from ._client import AzureAIClient, AzureAIProjectAgentOptions, RawAzureAIClient  # pyright: ignore[reportDeprecated]
+from ._deprecated_azure_openai import (
+    AzureOpenAIAssistantsClient,  # pyright: ignore[reportDeprecated]
+    AzureOpenAIAssistantsOptions,
+    AzureOpenAIChatClient,  # pyright: ignore[reportDeprecated]
+    AzureOpenAIChatOptions,
+    AzureOpenAIConfigMixin,
+    AzureOpenAIEmbeddingClient,  # pyright: ignore[reportDeprecated]
+    AzureOpenAIResponsesClient,  # pyright: ignore[reportDeprecated]
+    AzureOpenAIResponsesOptions,
+    AzureOpenAISettings,
+    AzureUserSecurityContext,
+)
 from ._embedding_client import (
    AzureAIInferenceEmbeddingClient,
    AzureAIInferenceEmbeddingOptions,
    AzureAIInferenceEmbeddingSettings,
    RawAzureAIInferenceEmbeddingClient,
 )
-from ._foundry_memory_provider import FoundryMemoryProvider
-from ._project_provider import AzureAIProjectAgentProvider
+from ._entra_id_authentication import AzureCredentialTypes, AzureTokenProvider
+from ._project_provider import AzureAIProjectAgentProvider  # pyright: ignore[reportDeprecated]
 from ._shared import AzureAISettings

 try:
    __version__ = importlib.metadata.version(__name__)
 except importlib.metadata.PackageNotFoundError:
-    __version__ = "0.0.0"  # Fallback for development mode
+    __version__ = "0.0.0"

 __all__ = [
    "AzureAIAgentClient",
@@ -31,7 +43,18 @@ __all__ = [
    "AzureAIProjectAgentOptions",
    "AzureAIProjectAgentProvider",
    "AzureAISettings",
-    "FoundryMemoryProvider",
+    "AzureCredentialTypes",
+    "AzureOpenAIAssistantsClient",
+    "AzureOpenAIAssistantsOptions",
+    "AzureOpenAIChatClient",
+    "AzureOpenAIChatOptions",
+    "AzureOpenAIConfigMixin",
+    "AzureOpenAIEmbeddingClient",
+    "AzureOpenAIResponsesClient",
+    "AzureOpenAIResponsesOptions",
+    "AzureOpenAISettings",
+    "AzureTokenProvider",
+    "AzureUserSecurityContext",
    "RawAzureAIClient",
    "RawAzureAIInferenceEmbeddingClient",
    "__version__",
@@ -18,19 +18,23 @@ from agent_framework import (
 from agent_framework._mcp import MCPTool
 from agent_framework._settings import load_settings
 from agent_framework._tools import ToolTypes
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
 from azure.ai.agents.aio import AgentsClient
 from azure.ai.agents.models import Agent as AzureAgent
 from azure.ai.agents.models import ResponseFormatJsonSchema, ResponseFormatJsonSchemaType
 from pydantic import BaseModel

-from ._chat_client import AzureAIAgentClient, AzureAIAgentOptions
+from ._chat_client import AzureAIAgentClient, AzureAIAgentOptions  # pyright: ignore[reportDeprecated]
+from ._entra_id_authentication import AzureCredentialTypes
 from ._shared import AzureAISettings, to_azure_ai_agent_tools

 if sys.version_info >= (3, 13):
    from typing import Self, TypeVar  # type: ignore # pragma: no cover
 else:
    from typing_extensions import Self, TypeVar  # type: ignore # pragma: no cover
+if sys.version_info >= (3, 13):
+    from warnings import deprecated  # type: ignore # pragma: no cover
+else:
+    from typing_extensions import deprecated  # type: ignore # pragma: no cover
 if sys.version_info >= (3, 11):
    from typing import TypedDict  # type: ignore # pragma: no cover
 else:
@@ -47,6 +51,11 @@ OptionsCoT = TypeVar(
 )


+@deprecated(
+    "AzureAIAgentClient and the AzureAIAgentsProvider are deprecated. "
+    "They target the V1 Agents Service API and have no direct replacement; "
+    "for new Foundry projects, use FoundryAgent."
+)
 class AzureAIAgentsProvider(Generic[OptionsCoT]):
    """Provider for Azure AI Agent Service V1 (Persistent Agents API).

@@ -426,7 +435,7 @@ class AzureAIAgentsProvider(Generic[OptionsCoT]):
            context_providers: Context providers to include during agent invocation.
        """
        # Create the underlying client
-        client = AzureAIAgentClient(
+        client = AzureAIAgentClient(  # pyright: ignore[reportDeprecated]
            agents_client=self._agents_client,
            agent_id=agent.id,
            agent_name=agent.name,
@@ -36,7 +36,6 @@ from agent_framework import (
 )
 from agent_framework._settings import load_settings
 from agent_framework._tools import ToolTypes
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
 from agent_framework.exceptions import (
    ChatClientException,
    ChatClientInvalidRequestException,
@@ -92,12 +91,14 @@ from azure.ai.agents.models import (
 )
 from pydantic import BaseModel

+from ._entra_id_authentication import AzureCredentialTypes
 from ._shared import AzureAISettings, resolve_file_ids, to_azure_ai_agent_tools

 if sys.version_info >= (3, 13):
    from typing import TypeVar  # type: ignore # pragma: no cover
+    from warnings import deprecated  # type: ignore # pragma: no cover
 else:
-    from typing_extensions import TypeVar  # type: ignore # pragma: no cover
+    from typing_extensions import TypeVar, deprecated  # type: ignore # pragma: no cover
 if sys.version_info >= (3, 12):
    from typing import override  # type: ignore # pragma: no cover
 else:
@@ -210,6 +211,11 @@ AzureAIAgentOptionsT = TypeVar(
 # endregion


+@deprecated(
+    "AzureAIAgentClient is deprecated. "
+    "It targets the V1 Agents Service API and has no direct replacement; "
+    "for new Foundry projects, use FoundryAgent."
+)
 class AzureAIAgentClient(
    FunctionInvocationLayer[AzureAIAgentOptionsT],
    ChatMiddlewareLayer[AzureAIAgentOptionsT],
@@ -221,7 +227,8 @@ class AzureAIAgentClient(

    .. deprecated::
        AzureAIAgentClient is deprecated and will be removed in a future release.
-        Use :class:`AzureAIClient` instead for the V2 (Projects/Responses) API.
+        It targets the V1 Agents Service API and has no direct replacement.
+        For new Foundry projects, use :class:`FoundryAgent`.
    """

    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai"  # type: ignore[reportIncompatibleVariableOverride, misc]
@@ -239,7 +246,8 @@ class AzureAIAgentClient(

        .. deprecated::
            This method is deprecated and will be removed in a future release.
-            Use :meth:`AzureAIClient.get_code_interpreter_tool` instead.
+            For new Foundry projects, configure hosted tools on the Foundry agent definition
+            in the service instead.

        Keyword Args:
            file_ids: List of uploaded file IDs or Content objects to make available to
@@ -272,7 +280,7 @@ class AzureAIAgentClient(
        """
        warnings.warn(
            "AzureAIAgentClient.get_code_interpreter_tool() is deprecated and will be removed in a future release; "
-            "use AzureAIClient.get_code_interpreter_tool() instead.",
+            "for new Foundry projects, configure hosted tools on the Foundry agent definition in the service instead.",
            DeprecationWarning,
            stacklevel=2,
        )
@@ -288,7 +296,8 @@ class AzureAIAgentClient(

        .. deprecated::
            This method is deprecated and will be removed in a future release.
-            Use :meth:`AzureAIClient.get_file_search_tool` instead.
+            For new Foundry projects, configure hosted tools on the Foundry agent definition
+            in the service instead.

        Keyword Args:
            vector_store_ids: List of vector store IDs to search within.
@@ -308,7 +317,7 @@ class AzureAIAgentClient(
        """
        warnings.warn(
            "AzureAIAgentClient.get_file_search_tool() is deprecated and will be removed in a future release; "
-            "use AzureAIClient.get_file_search_tool() instead.",
+            "for new Foundry projects, configure hosted tools on the Foundry agent definition in the service instead.",
            DeprecationWarning,
            stacklevel=2,
        )
@@ -325,7 +334,8 @@ class AzureAIAgentClient(

        .. deprecated::
            This method is deprecated and will be removed in a future release.
-            Use :meth:`AzureAIClient.get_web_search_tool` instead.
+            For new Foundry projects, configure hosted tools on the Foundry agent definition
+            in the service instead.

        For Azure AI Agents, web search uses Bing Grounding or Bing Custom Search.
        If no arguments are provided, attempts to read from environment variables.
@@ -369,7 +379,7 @@ class AzureAIAgentClient(
        """
        warnings.warn(
            "AzureAIAgentClient.get_web_search_tool() is deprecated and will be removed in a future release; "
-            "use AzureAIClient.get_web_search_tool() instead.",
+            "for new Foundry projects, configure hosted tools on the Foundry agent definition in the service instead.",
            DeprecationWarning,
            stacklevel=2,
        )
@@ -410,7 +420,8 @@ class AzureAIAgentClient(

        .. deprecated::
            This method is deprecated and will be removed in a future release.
-            Use :meth:`AzureAIClient.get_mcp_tool` instead.
+            For new Foundry projects, configure hosted tools on the Foundry agent definition
+            in the service instead.

        This configures an MCP (Model Context Protocol) server that will be called
        by Azure AI's service. The tools from this MCP server are executed remotely
@@ -446,7 +457,7 @@ class AzureAIAgentClient(
        """
        warnings.warn(
            "AzureAIAgentClient.get_mcp_tool() is deprecated and will be removed in a future release; "
-            "use AzureAIClient.get_mcp_tool() instead.",
+            "for new Foundry projects, configure hosted tools on the Foundry agent definition in the service instead.",
            DeprecationWarning,
            stacklevel=2,
        )
@@ -561,12 +572,6 @@ class AzureAIAgentClient(
                client: AzureAIAgentClient[MyOptions] = AzureAIAgentClient(credential=credential)
                response = await client.get_response("Hello", options={"my_custom_option": "value"})
        """
-        warnings.warn(
-            "AzureAIAgentClient is deprecated and will be removed in a future release; "
-            "use AzureAIClient instead for the V2 (Projects/Responses) API.",
-            DeprecationWarning,
-            stacklevel=2,
-        )
        azure_ai_settings = load_settings(
            AzureAISettings,
            env_prefix="AZURE_AI_",
@@ -30,10 +30,9 @@ from agent_framework import (
 )
 from agent_framework._settings import load_settings
 from agent_framework._tools import ToolTypes
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
 from agent_framework.observability import ChatTelemetryLayer
 from agent_framework.openai import OpenAIResponsesOptions
-from agent_framework.openai._responses_client import RawOpenAIResponsesClient
+from agent_framework_openai._chat_client import RawOpenAIChatClient
 from azure.ai.projects.aio import AIProjectClient
 from azure.ai.projects.models import (
    ApproximateLocation,
@@ -50,12 +49,14 @@ from azure.ai.projects.models import (
 from azure.ai.projects.models import FileSearchTool as ProjectsFileSearchTool
 from azure.core.exceptions import ResourceNotFoundError

+from ._entra_id_authentication import AzureCredentialTypes
 from ._shared import AzureAISettings, create_text_format_config, resolve_file_ids

 if sys.version_info >= (3, 13):
    from typing import TypeVar  # type: ignore # pragma: no cover
+    from warnings import deprecated  # type: ignore # pragma: no cover
 else:
-    from typing_extensions import TypeVar  # type: ignore # pragma: no cover
+    from typing_extensions import TypeVar, deprecated  # type: ignore # pragma: no cover
 if sys.version_info >= (3, 12):
    from typing import override  # type: ignore # pragma: no cover
 else:
@@ -68,7 +69,7 @@ else:
 logger = logging.getLogger("agent_framework.azure")


-class AzureAIProjectAgentOptions(OpenAIResponsesOptions, total=False):
+class AzureAIProjectAgentOptions(OpenAIResponsesOptions, total=False):  # type: ignore[misc, call-arg]
    """Azure AI Project Agent options."""

    rai_config: RaiConfig
@@ -88,8 +89,13 @@ AzureAIClientOptionsT = TypeVar(
 _DOC_INDEX_PATTERN = re.compile(r"doc_(\d+)")


-class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[AzureAIClientOptionsT]):
-    """Raw Azure AI client without middleware, telemetry, or function invocation layers.
+@deprecated(
+    "RawAzureAIClient is deprecated. "
+    "Use RawFoundryAgentChatClient for low-level Foundry agent client customization, "
+    "or FoundryAgent for the recommended production API."
+)
+class RawAzureAIClient(RawOpenAIChatClient[AzureAIClientOptionsT], Generic[AzureAIClientOptionsT]):
+    """Deprecated raw Azure AI client without middleware, telemetry, or function invocation layers.

    Warning:
        **This class should not normally be used directly.** It does not include middleware,
@@ -101,7 +107,8 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[
        2. **ChatMiddlewareLayer** - Applies chat middleware per model call and stays outside telemetry
        3. **ChatTelemetryLayer** - Must stay inside chat middleware for correct per-call telemetry

-        Use ``AzureAIClient`` instead for a fully-featured client with all layers applied.
+        Use ``RawFoundryAgentChatClient`` for low-level Foundry agent customization, or
+        ``FoundryAgent`` for the recommended production API.
    """

    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai"  # type: ignore[reportIncompatibleVariableOverride, misc]
@@ -215,8 +222,10 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[
            project_client = AIProjectClient(**project_client_kwargs)
            should_close_client = True

-        # Initialize parent
-        super().__init__(
+        # Initialize parent with OpenAI client from project
+        super().__init__(  # type: ignore
+            async_client=project_client.get_openai_client(),
+            model=azure_ai_settings.get("model"),  # type: ignore[arg-type]
            additional_properties=additional_properties,
        )

@@ -680,10 +689,6 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[

        return result, instructions

-    async def _initialize_client(self) -> None:
-        """Initialize OpenAI client."""
-        self.client = self.project_client.get_openai_client()  # type: ignore
-
    def _update_agent_name_and_description(self, agent_name: str | None, description: str | None = None) -> None:
        """Update the agent name in the chat client.

@@ -842,7 +847,7 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[
        if not stream:

            async def _enrich_response() -> ChatResponse:
-                response = await super(RawAzureAIClient, self)._inner_get_response(
+                response = await super(RawAzureAIClient, self)._inner_get_response(  # pyright: ignore[reportDeprecated]
                    messages=messages, options=options, stream=False, **kwargs
                )
                get_urls = self._extract_azure_search_urls(response.raw_representation.output)  # type: ignore[union-attr]
@@ -1182,8 +1187,8 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[
        It does NOT create an agent on the Azure AI service - the actual agent
        will be created on the server during the first invocation (run).

-        For creating and managing persistent agents on the server, use
-        :class:`~agent_framework_azure_ai.AzureAIProjectAgentProvider` instead.
+        For working with pre-configured persistent agents on the server, use
+        :class:`~agent_framework_azure_ai.FoundryAgent` instead.

        Keyword Args:
            id: The unique identifier for the agent. Will be created automatically if not provided.
@@ -1213,21 +1218,23 @@ class RawAzureAIClient(RawOpenAIResponsesClient[AzureAIClientOptionsT], Generic[
        )


+@deprecated("AzureAIClient is deprecated. Use FoundryAgent instead.")
 class AzureAIClient(
    FunctionInvocationLayer[AzureAIClientOptionsT],
    ChatMiddlewareLayer[AzureAIClientOptionsT],
    ChatTelemetryLayer[AzureAIClientOptionsT],
-    RawAzureAIClient[AzureAIClientOptionsT],
+    RawAzureAIClient[AzureAIClientOptionsT],  # pyright: ignore[reportDeprecated]
    Generic[AzureAIClientOptionsT],
 ):
-    """Azure AI client with middleware, telemetry, and function invocation support.
+    """Deprecated Azure AI client with middleware, telemetry, and function invocation support.

-    This is the recommended client for most use cases. It includes:
+    This class is deprecated. Use ``FoundryAgent`` instead for connecting to
+    pre-configured agents in Foundry. It includes:
    - Chat middleware support for request/response interception
    - OpenTelemetry-based telemetry for observability
    - Automatic function/tool invocation handling

-    For a minimal implementation without these features, use :class:`RawAzureAIClient`.
+    For a minimal implementation without these features, use :class:`RawFoundryAgentChatClient`.
    """

    def __init__(
@@ -0,0 +1,897 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+"""Deprecated Azure OpenAI client classes.
+
+All classes in this module are deprecated and will be removed in a future release.
+Migrate to the ``agent_framework_openai`` package equivalents with an ``AsyncAzureOpenAI`` client,
+or use ``FoundryChatClient`` for Azure AI Foundry projects.
+"""
+
+from __future__ import annotations
+
+import json
+import logging
+import sys
+from collections.abc import Mapping, Sequence
+from copy import copy
+from typing import TYPE_CHECKING, Any, ClassVar, Final, Generic, cast
+from urllib.parse import urljoin, urlparse
+
+from agent_framework._middleware import ChatMiddlewareLayer
+from agent_framework._settings import SecretString, load_settings
+from agent_framework._telemetry import AGENT_FRAMEWORK_USER_AGENT, APP_INFO, prepend_agent_framework_to_user_agent
+from agent_framework._tools import FunctionInvocationConfiguration, FunctionInvocationLayer
+from agent_framework._types import Annotation, Content
+from agent_framework.observability import ChatTelemetryLayer, EmbeddingTelemetryLayer
+from agent_framework_openai._assistants_client import OpenAIAssistantsClient, OpenAIAssistantsOptions
+from agent_framework_openai._chat_client import OpenAIChatOptions, RawOpenAIChatClient
+from agent_framework_openai._chat_completion_client import OpenAIChatCompletionOptions, RawOpenAIChatCompletionClient
+from agent_framework_openai._embedding_client import OpenAIEmbeddingOptions, RawOpenAIEmbeddingClient
+from agent_framework_openai._shared import OpenAIBase
+from azure.ai.projects.aio import AIProjectClient
+from openai import AsyncOpenAI
+from openai.lib.azure import AsyncAzureOpenAI
+from pydantic import BaseModel
+
+from ._entra_id_authentication import AzureCredentialTypes, AzureTokenProvider, resolve_credential_to_token_provider
+
+if sys.version_info >= (3, 13):
+    from typing import TypeVar  # type: ignore # pragma: no cover
+    from warnings import deprecated  # type: ignore # pragma: no cover
+else:
+    from typing_extensions import TypeVar, deprecated  # type: ignore # pragma: no cover
+if sys.version_info >= (3, 12):
+    from typing import override  # type: ignore # pragma: no cover
+else:
+    from typing_extensions import override  # type: ignore # pragma: no cover
+if sys.version_info >= (3, 11):
+    from typing import TypedDict  # type: ignore # pragma: no cover
+else:
+    from typing_extensions import TypedDict  # type: ignore # pragma: no cover
+
+if TYPE_CHECKING:
+    from agent_framework._middleware import MiddlewareTypes
+    from openai.types.chat.chat_completion import Choice
+    from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
+
+logger: logging.Logger = logging.getLogger(__name__)
+
+
+# region Constants and Settings
+
+DEFAULT_AZURE_API_VERSION: Final[str] = "2024-10-21"
+DEFAULT_AZURE_TOKEN_ENDPOINT: Final[str] = "https://cognitiveservices.azure.com/.default"  # noqa: S105
+
+
+class AzureOpenAISettings(TypedDict, total=False):
+    """AzureOpenAI model settings.
+
+    Settings are resolved in this order: explicit keyword arguments, values from an
+    explicitly provided .env file, then environment variables with the prefix
+    'AZURE_OPENAI_'. If settings are missing after resolution, validation will fail.
+
+    Keyword Args:
+        endpoint: The endpoint of the Azure deployment.
+        chat_deployment_name: The name of the Azure Chat deployment.
+        responses_deployment_name: The name of the Azure Responses deployment.
+        embedding_deployment_name: The name of the Azure Embedding deployment.
+        api_key: The API key for the Azure deployment.
+        api_version: The API version to use.
+        base_url: The url of the Azure deployment.
+        token_endpoint: The token endpoint to use to retrieve the authentication token.
+    """
+
+    chat_deployment_name: str | None
+    responses_deployment_name: str | None
+    embedding_deployment_name: str | None
+    endpoint: str | None
+    base_url: str | None
+    api_key: SecretString | None
+    api_version: str | None
+    token_endpoint: str | None
+
+
+def _apply_azure_defaults(
+    settings: AzureOpenAISettings,
+    default_api_version: str = DEFAULT_AZURE_API_VERSION,
+    default_token_endpoint: str = DEFAULT_AZURE_TOKEN_ENDPOINT,
+) -> None:
+    """Apply default values for api_version and token_endpoint after loading settings.
+
+    Args:
+        settings: The loaded Azure OpenAI settings dict.
+        default_api_version: The default API version to use if not set.
+        default_token_endpoint: The default token endpoint to use if not set.
+    """
+    if not settings.get("api_version"):
+        settings["api_version"] = default_api_version
+    if not settings.get("token_endpoint"):
+        settings["token_endpoint"] = default_token_endpoint
+
+
+# endregion
+
+
+# region AzureOpenAIConfigMixin
+
+
+class AzureOpenAIConfigMixin(OpenAIBase):
+    """Internal class for configuring a connection to an Azure OpenAI service."""
+
+    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai.openai"
+
+    def __init__(
+        self,
+        deployment_name: str,
+        endpoint: str | None = None,
+        base_url: str | None = None,
+        api_version: str = DEFAULT_AZURE_API_VERSION,
+        api_key: str | None = None,
+        token_endpoint: str | None = None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None = None,
+        default_headers: Mapping[str, str] | None = None,
+        client: AsyncOpenAI | None = None,
+        instruction_role: str | None = None,
+        **kwargs: Any,
+    ) -> None:
+        """Configure a connection to an Azure OpenAI service.
+
+        Args:
+            deployment_name: Name of the deployment.
+            endpoint: The specific endpoint URL for the deployment.
+            base_url: The base URL for Azure services.
+            api_version: Azure API version.
+            api_key: API key for Azure services.
+            token_endpoint: Azure AD token scope.
+            credential: Azure credential or token provider for authentication.
+            default_headers: Default headers for HTTP requests.
+            client: An existing client to use.
+            instruction_role: The role to use for 'instruction' messages.
+            kwargs: Additional keyword arguments.
+        """
+        merged_headers = dict(copy(default_headers)) if default_headers else {}
+        if APP_INFO:
+            merged_headers.update(APP_INFO)
+            merged_headers = prepend_agent_framework_to_user_agent(merged_headers)
+        if not client:
+            ad_token_provider = None
+            if not api_key and credential:
+                ad_token_provider = resolve_credential_to_token_provider(credential, token_endpoint)
+
+            if not api_key and not ad_token_provider:
+                raise ValueError("Please provide either api_key, credential, or a client.")
+
+            if not endpoint and not base_url:
+                raise ValueError("Please provide an endpoint or a base_url")
+
+            args: dict[str, Any] = {
+                "default_headers": merged_headers,
+            }
+            if api_version:
+                args["api_version"] = api_version
+            if ad_token_provider:
+                args["azure_ad_token_provider"] = ad_token_provider
+            if api_key:
+                args["api_key"] = api_key
+            if base_url:
+                args["base_url"] = str(base_url)
+            if endpoint and not base_url:
+                args["azure_endpoint"] = str(endpoint)
+            if deployment_name:
+                args["azure_deployment"] = deployment_name
+            if "websocket_base_url" in kwargs:
+                args["websocket_base_url"] = kwargs.pop("websocket_base_url")
+
+            client = AsyncAzureOpenAI(**args)
+
+        self.endpoint = str(endpoint)
+        self.base_url = str(base_url)
+        self.api_version = api_version
+        self.deployment_name = deployment_name
+        self.instruction_role = instruction_role
+        if default_headers:
+            from agent_framework._telemetry import USER_AGENT_KEY
+
+            def_headers = {k: v for k, v in default_headers.items() if k != USER_AGENT_KEY}
+        else:
+            def_headers = None
+        self.default_headers = def_headers
+
+        super().__init__(model_id=deployment_name, client=client, **kwargs)
+
+
+# endregion
+
+
+# region AzureOpenAIResponsesClient
+
+
+AzureOpenAIResponsesOptionsT = TypeVar(
+    "AzureOpenAIResponsesOptionsT",
+    bound=TypedDict,  # type: ignore[valid-type]
+    default="OpenAIChatOptions",
+    covariant=True,
+)
+
+AzureOpenAIResponsesOptions = OpenAIChatOptions
+
+
+@deprecated(
+    "AzureOpenAIResponsesClient is deprecated. "
+    "Use OpenAIChatClient with an AsyncAzureOpenAI client, or FoundryChatClient for Foundry projects."
+)
+class AzureOpenAIResponsesClient(  # type: ignore[misc]
+    FunctionInvocationLayer[AzureOpenAIResponsesOptionsT],
+    ChatMiddlewareLayer[AzureOpenAIResponsesOptionsT],
+    ChatTelemetryLayer[AzureOpenAIResponsesOptionsT],
+    RawOpenAIChatClient[AzureOpenAIResponsesOptionsT],
+    Generic[AzureOpenAIResponsesOptionsT],
+):
+    """Deprecated Azure Responses client. Use OpenAIChatClient with an AsyncAzureOpenAI client instead."""
+
+    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai.openai"
+
+    def __init__(
+        self,
+        *,
+        api_key: str | None = None,
+        deployment_name: str | None = None,
+        endpoint: str | None = None,
+        base_url: str | None = None,
+        api_version: str | None = None,
+        token_endpoint: str | None = None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None = None,
+        default_headers: Mapping[str, str] | None = None,
+        async_client: AsyncOpenAI | None = None,
+        project_client: Any | None = None,
+        project_endpoint: str | None = None,
+        allow_preview: bool | None = None,
+        env_file_path: str | None = None,
+        env_file_encoding: str | None = None,
+        instruction_role: str | None = None,
+        middleware: Sequence[MiddlewareTypes] | None = None,
+        function_invocation_configuration: FunctionInvocationConfiguration | None = None,
+        **kwargs: Any,
+    ) -> None:
+        """Initialize an Azure OpenAI Responses client.
+
+        Keyword Args:
+            api_key: The API key.
+            deployment_name: The deployment name.
+            endpoint: The deployment endpoint.
+            base_url: The deployment base URL.
+            api_version: The deployment API version.
+            token_endpoint: The token endpoint to request an Azure token.
+            credential: Azure credential or token provider for authentication.
+            default_headers: Default headers for HTTP requests.
+            async_client: An existing client to use.
+            project_client: An existing AIProjectClient to use.
+            project_endpoint: The Azure AI Foundry project endpoint URL.
+            allow_preview: Enables preview opt-in on internally-created AIProjectClient.
+            env_file_path: Path to .env file for settings.
+            env_file_encoding: Encoding for .env file.
+            instruction_role: The role to use for 'instruction' messages.
+            middleware: Optional sequence of middleware.
+            function_invocation_configuration: Optional function invocation configuration.
+            kwargs: Additional keyword arguments.
+        """
+        if (model_id := kwargs.pop("model_id", None)) and not deployment_name:
+            deployment_name = str(model_id)
+
+        if async_client is None and (project_client is not None or project_endpoint is not None):
+            async_client = self._create_client_from_project(
+                project_client=project_client,
+                project_endpoint=project_endpoint,
+                credential=credential,
+                allow_preview=allow_preview,
+            )
+
+        azure_openai_settings = load_settings(
+            AzureOpenAISettings,
+            env_prefix="AZURE_OPENAI_",
+            api_key=api_key,
+            base_url=base_url,
+            endpoint=endpoint,
+            responses_deployment_name=deployment_name,
+            api_version=api_version,
+            env_file_path=env_file_path,
+            env_file_encoding=env_file_encoding,
+            token_endpoint=token_endpoint,
+        )
+        _apply_azure_defaults(azure_openai_settings, default_api_version="preview")
+        endpoint_value = azure_openai_settings.get("endpoint")
+        if (
+            not azure_openai_settings.get("base_url")
+            and endpoint_value
+            and (hostname := urlparse(str(endpoint_value)).hostname)
+            and hostname.endswith(".openai.azure.com")
+        ):
+            azure_openai_settings["base_url"] = urljoin(str(endpoint_value), "/openai/v1/")
+
+        responses_deployment_name = azure_openai_settings.get("responses_deployment_name")
+        if not responses_deployment_name:
+            raise ValueError(
+                "Azure OpenAI deployment name is required. Set via 'deployment_name' parameter "
+                "or 'AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME' environment variable."
+            )
+
+        if not async_client:
+            # Create the Azure OpenAI client directly
+            merged_headers = dict(copy(default_headers)) if default_headers else {}
+            if APP_INFO:
+                merged_headers.update(APP_INFO)
+                merged_headers = prepend_agent_framework_to_user_agent(merged_headers)
+
+            api_key_secret = azure_openai_settings.get("api_key")
+            ad_token_provider = None
+            if not api_key_secret and credential:
+                ad_token_provider = resolve_credential_to_token_provider(
+                    credential, azure_openai_settings.get("token_endpoint")
+                )
+
+            if not api_key_secret and not ad_token_provider:
+                raise ValueError("Please provide either api_key, credential, or a client.")
+
+            client_endpoint = azure_openai_settings.get("endpoint")
+            client_base_url = azure_openai_settings.get("base_url")
+            if not client_endpoint and not client_base_url:
+                raise ValueError("Please provide an endpoint or a base_url")
+
+            client_args: dict[str, Any] = {"default_headers": merged_headers}
+            if resolved_api_version := azure_openai_settings.get("api_version"):
+                client_args["api_version"] = resolved_api_version
+            if ad_token_provider:
+                client_args["azure_ad_token_provider"] = ad_token_provider
+            if api_key_secret:
+                client_args["api_key"] = api_key_secret.get_secret_value()
+            if client_base_url:
+                client_args["base_url"] = str(client_base_url)
+            if client_endpoint and not client_base_url:
+                client_args["azure_endpoint"] = str(client_endpoint)
+            if responses_deployment_name:
+                client_args["azure_deployment"] = responses_deployment_name
+            if "websocket_base_url" in kwargs:
+                client_args["websocket_base_url"] = kwargs.pop("websocket_base_url")
+
+            async_client = AsyncAzureOpenAI(**client_args)
+
+        # Store Azure-specific attributes for serialization
+        self.endpoint = str(endpoint_value) if endpoint_value else None
+        self.api_version = azure_openai_settings.get("api_version") or ""
+        self.deployment_name = responses_deployment_name
+
+        super().__init__(
+            async_client=async_client,
+            model=responses_deployment_name,
+            api_version=azure_openai_settings.get("api_version"),
+            instruction_role=instruction_role,
+            default_headers=default_headers,
+            middleware=middleware,  # type: ignore[arg-type]
+            function_invocation_configuration=function_invocation_configuration,
+            **kwargs,
+        )
+
+    @staticmethod
+    def _create_client_from_project(
+        *,
+        project_client: AIProjectClient | None,
+        project_endpoint: str | None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None,
+        allow_preview: bool | None = None,
+    ) -> AsyncOpenAI:
+        """Create an AsyncOpenAI client from an Azure AI Foundry project."""
+        if project_client is not None:
+            return project_client.get_openai_client()
+
+        if not project_endpoint:
+            raise ValueError("Azure AI project endpoint is required when project_client is not provided.")
+        if not credential:
+            raise ValueError("Azure credential is required when using project_endpoint without a project_client.")
+        project_client_kwargs: dict[str, Any] = {
+            "endpoint": project_endpoint,
+            "credential": credential,  # type: ignore[arg-type]
+            "user_agent": AGENT_FRAMEWORK_USER_AGENT,
+        }
+        if allow_preview is not None:
+            project_client_kwargs["allow_preview"] = allow_preview
+        project_client = AIProjectClient(**project_client_kwargs)
+        return project_client.get_openai_client()
+
+    @override
+    def _check_model_presence(self, options: dict[str, Any]) -> None:
+        if not options.get("model"):
+            if not self.model:
+                raise ValueError("deployment_name must be a non-empty string")
+            options["model"] = self.model
+
+
+# endregion
+
+
+# region AzureOpenAIChatClient
+
+
+ResponseModelT = TypeVar("ResponseModelT", bound=BaseModel | None, default=None)
+
+
+class AzureUserSecurityContext(TypedDict, total=False):
+    """User security context for Azure AI applications.
+
+    These fields help security operations teams investigate and mitigate security
+    incidents by providing context about the application and end user.
+    """
+
+    application_name: str
+    """Name of the application making the request."""
+
+    end_user_id: str
+    """Unique identifier for the end user (recommend hashing username/email)."""
+
+    end_user_tenant_id: str
+    """Microsoft 365 tenant ID the end user belongs to. Required for multi-tenant apps."""
+
+    source_ip: str
+    """The original client's IP address."""
+
+
+class AzureOpenAIChatOptions(OpenAIChatCompletionOptions[ResponseModelT], Generic[ResponseModelT], total=False):
+    """Azure OpenAI-specific chat options dict.
+
+    Extends OpenAIChatCompletionOptions with Azure-specific options including
+    the "On Your Data" feature and enhanced security context.
+    """
+
+    data_sources: list[dict[str, Any]]
+    """Azure "On Your Data" data sources for retrieval-augmented generation."""
+
+    user_security_context: AzureUserSecurityContext
+    """Enhanced security context for Azure Defender integration."""
+
+    n: int
+    """Number of chat completion choices to generate for each input message."""
+
+
+AzureOpenAIChatOptionsT = TypeVar(
+    "AzureOpenAIChatOptionsT",
+    bound=TypedDict,  # type: ignore[valid-type]
+    default="AzureOpenAIChatOptions",
+    covariant=True,
+)
+
+
+@deprecated("AzureOpenAIChatClient is deprecated. Use OpenAIChatCompletionClient with an AsyncAzureOpenAI client.")
+class AzureOpenAIChatClient(  # type: ignore[misc]
+    FunctionInvocationLayer[AzureOpenAIChatOptionsT],
+    ChatMiddlewareLayer[AzureOpenAIChatOptionsT],
+    ChatTelemetryLayer[AzureOpenAIChatOptionsT],
+    RawOpenAIChatCompletionClient[AzureOpenAIChatOptionsT],
+    Generic[AzureOpenAIChatOptionsT],
+):
+    """Deprecated Azure OpenAI Chat client. Use OpenAIChatCompletionClient with AsyncAzureOpenAI instead."""
+
+    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai.openai"
+
+    def __init__(
+        self,
+        *,
+        api_key: str | None = None,
+        deployment_name: str | None = None,
+        endpoint: str | None = None,
+        base_url: str | None = None,
+        api_version: str | None = None,
+        token_endpoint: str | None = None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None = None,
+        default_headers: Mapping[str, str] | None = None,
+        async_client: AsyncAzureOpenAI | None = None,
+        additional_properties: dict[str, Any] | None = None,
+        env_file_path: str | None = None,
+        env_file_encoding: str | None = None,
+        instruction_role: str | None = None,
+        middleware: Sequence[MiddlewareTypes] | None = None,
+        function_invocation_configuration: FunctionInvocationConfiguration | None = None,
+    ) -> None:
+        """Initialize an Azure OpenAI Chat completion client.
+
+        Keyword Args:
+            api_key: The API key.
+            deployment_name: The deployment name.
+            endpoint: The deployment endpoint.
+            base_url: The deployment base URL.
+            api_version: The deployment API version.
+            token_endpoint: The token endpoint to request an Azure token.
+            credential: Azure credential or token provider for authentication.
+            default_headers: Default headers for HTTP requests.
+            async_client: An existing client to use.
+            additional_properties: Additional properties stored on the client instance.
+            env_file_path: Path to .env file for settings.
+            env_file_encoding: Encoding for .env file.
+            instruction_role: The role to use for 'instruction' messages.
+            middleware: Optional sequence of middleware.
+            function_invocation_configuration: Optional function invocation configuration.
+        """
+        azure_openai_settings = load_settings(
+            AzureOpenAISettings,
+            env_prefix="AZURE_OPENAI_",
+            api_key=api_key,
+            base_url=base_url,
+            endpoint=endpoint,
+            chat_deployment_name=deployment_name,
+            api_version=api_version,
+            env_file_path=env_file_path,
+            env_file_encoding=env_file_encoding,
+            token_endpoint=token_endpoint,
+        )
+        _apply_azure_defaults(azure_openai_settings)
+
+        chat_deployment_name = azure_openai_settings.get("chat_deployment_name")
+        if not chat_deployment_name:
+            raise ValueError(
+                "Azure OpenAI deployment name is required. Set via 'deployment_name' parameter "
+                "or 'AZURE_OPENAI_CHAT_DEPLOYMENT_NAME' environment variable."
+            )
+
+        if not async_client:
+            # Create the Azure OpenAI client directly
+            merged_headers = dict(copy(default_headers)) if default_headers else {}
+            if APP_INFO:
+                merged_headers.update(APP_INFO)
+                merged_headers = prepend_agent_framework_to_user_agent(merged_headers)
+
+            api_key_secret = azure_openai_settings.get("api_key")
+            ad_token_provider = None
+            if not api_key_secret and credential:
+                ad_token_provider = resolve_credential_to_token_provider(
+                    credential, azure_openai_settings.get("token_endpoint")
+                )
+
+            if not api_key_secret and not ad_token_provider:
+                raise ValueError("Please provide either api_key, credential, or a client.")
+
+            endpoint_value = azure_openai_settings.get("endpoint")
+            base_url_value = azure_openai_settings.get("base_url")
+            if not endpoint_value and not base_url_value:
+                raise ValueError("Please provide an endpoint or a base_url")
+
+            client_args: dict[str, Any] = {"default_headers": merged_headers}
+            if resolved_api_version := azure_openai_settings.get("api_version"):
+                client_args["api_version"] = resolved_api_version
+            if ad_token_provider:
+                client_args["azure_ad_token_provider"] = ad_token_provider
+            if api_key_secret:
+                client_args["api_key"] = api_key_secret.get_secret_value()
+            if base_url_value:
+                client_args["base_url"] = str(base_url_value)
+            if endpoint_value and not base_url_value:
+                client_args["azure_endpoint"] = str(endpoint_value)
+            if chat_deployment_name:
+                client_args["azure_deployment"] = chat_deployment_name
+
+            async_client = AsyncAzureOpenAI(**client_args)
+
+        # Store Azure-specific attributes for serialization
+        self.endpoint = str(azure_openai_settings.get("endpoint") or "")
+        self.api_version = azure_openai_settings.get("api_version") or ""
+        self.deployment_name = chat_deployment_name
+
+        super().__init__(
+            async_client=async_client,
+            model=chat_deployment_name,
+            api_version=azure_openai_settings.get("api_version"),
+            instruction_role=instruction_role,
+            default_headers=default_headers,
+            additional_properties=additional_properties,
+            middleware=middleware,  # type: ignore[arg-type]
+            function_invocation_configuration=function_invocation_configuration,
+        )
+
+    @override
+    def _parse_text_from_openai(self, choice: Choice | ChunkChoice) -> Content | None:
+        """Parse the choice into a Content object with type='text'.
+
+        Overwritten from RawOpenAIChatCompletionClient to deal with Azure On Your Data function.
+        """
+        message = getattr(choice, "message", None)
+        if message is None:
+            message = getattr(choice, "delta", None)
+        if message is None:  # type: ignore
+            return None
+        if hasattr(message, "refusal") and message.refusal:
+            return Content.from_text(text=message.refusal, raw_representation=choice)
+        if not message.content:
+            return None
+        text_content = Content.from_text(text=message.content, raw_representation=choice)
+        if not message.model_extra or "context" not in message.model_extra:
+            return text_content
+
+        context_raw: object = cast(object, message.context)  # type: ignore[union-attr]
+        if isinstance(context_raw, str):
+            try:
+                context_raw = json.loads(context_raw)
+            except json.JSONDecodeError:
+                logger.warning("Context is not a valid JSON string, ignoring context.")
+                return text_content
+        if not isinstance(context_raw, dict):
+            logger.warning("Context is not a valid dictionary, ignoring context.")
+            return text_content
+        context = cast(dict[str, Any], context_raw)
+        if intent := context.get("intent"):
+            text_content.additional_properties = {"intent": intent}
+        citations = context.get("citations")
+        if isinstance(citations, list) and citations:
+            annotations: list[Annotation] = []
+            for citation_raw in cast(list[object], citations):
+                if not isinstance(citation_raw, dict):
+                    continue
+                citation = cast(dict[str, Any], citation_raw)
+                annotations.append(
+                    Annotation(
+                        type="citation",
+                        title=citation.get("title", ""),
+                        url=citation.get("url", ""),
+                        snippet=citation.get("content", ""),
+                        file_id=citation.get("filepath", ""),
+                        tool_name="Azure-on-your-Data",
+                        additional_properties={"chunk_id": citation.get("chunk_id", "")},
+                        raw_representation=citation,
+                    )
+                )
+            text_content.annotations = annotations
+        return text_content
+
+
+# endregion
+
+
+# region AzureOpenAIAssistantsClient
+
+
+AzureOpenAIAssistantsOptionsT = TypeVar(
+    "AzureOpenAIAssistantsOptionsT",
+    bound=TypedDict,  # type: ignore[valid-type]
+    default="OpenAIAssistantsOptions",
+    covariant=True,
+)
+
+AzureOpenAIAssistantsOptions = OpenAIAssistantsOptions
+
+
+@deprecated(
+    "AzureOpenAIAssistantsClient is deprecated. "
+    "Use OpenAIAssistantsClient (also deprecated) or migrate to OpenAIChatClient."
+)
+class AzureOpenAIAssistantsClient(
+    OpenAIAssistantsClient[AzureOpenAIAssistantsOptionsT], Generic[AzureOpenAIAssistantsOptionsT]
+):
+    """Deprecated Azure OpenAI Assistants client. Use OpenAIAssistantsClient or migrate to OpenAIChatClient."""
+
+    DEFAULT_AZURE_API_VERSION: ClassVar[str] = "2024-05-01-preview"
+
+    def __init__(
+        self,
+        *,
+        deployment_name: str | None = None,
+        assistant_id: str | None = None,
+        assistant_name: str | None = None,
+        assistant_description: str | None = None,
+        thread_id: str | None = None,
+        api_key: str | None = None,
+        endpoint: str | None = None,
+        base_url: str | None = None,
+        api_version: str | None = None,
+        token_endpoint: str | None = None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None = None,
+        default_headers: Mapping[str, str] | None = None,
+        async_client: AsyncAzureOpenAI | None = None,
+        env_file_path: str | None = None,
+        env_file_encoding: str | None = None,
+    ) -> None:
+        """Initialize an Azure OpenAI Assistants client.
+
+        Keyword Args:
+            deployment_name: The Azure OpenAI deployment name.
+            assistant_id: The ID of an Azure OpenAI assistant to use.
+            assistant_name: The name to use when creating new assistants.
+            assistant_description: The description to use when creating new assistants.
+            thread_id: Default thread ID to use for conversations.
+            api_key: The API key to use.
+            endpoint: The deployment endpoint.
+            base_url: The deployment base URL.
+            api_version: The deployment API version.
+            token_endpoint: The token endpoint to request an Azure token.
+            credential: Azure credential or token provider for authentication.
+            default_headers: Default headers for HTTP requests.
+            async_client: An existing client to use.
+            env_file_path: Path to .env file for settings.
+            env_file_encoding: Encoding for .env file.
+        """
+        azure_openai_settings = load_settings(
+            AzureOpenAISettings,
+            env_prefix="AZURE_OPENAI_",
+            api_key=api_key,
+            base_url=base_url,
+            endpoint=endpoint,
+            chat_deployment_name=deployment_name,
+            api_version=api_version,
+            env_file_path=env_file_path,
+            env_file_encoding=env_file_encoding,
+            token_endpoint=token_endpoint,
+        )
+        _apply_azure_defaults(azure_openai_settings, default_api_version=self.DEFAULT_AZURE_API_VERSION)
+
+        chat_deployment_name = azure_openai_settings.get("chat_deployment_name")
+        if not chat_deployment_name:
+            raise ValueError(
+                "Azure OpenAI deployment name is required. Set via 'deployment_name' parameter "
+                "or 'AZURE_OPENAI_CHAT_DEPLOYMENT_NAME' environment variable."
+            )
+
+        api_key_secret = azure_openai_settings.get("api_key")
+        token_scope = azure_openai_settings.get("token_endpoint")
+
+        ad_token_provider = None
+        if not async_client and not api_key_secret and credential:
+            ad_token_provider = resolve_credential_to_token_provider(credential, token_scope)
+
+        if not async_client and not api_key_secret and not ad_token_provider:
+            raise ValueError("Please provide either api_key, credential, or a client.")
+
+        if not async_client:
+            client_params: dict[str, Any] = {
+                "default_headers": default_headers,
+            }
+            if resolved_api_version := azure_openai_settings.get("api_version"):
+                client_params["api_version"] = resolved_api_version
+
+            if api_key_secret:
+                client_params["api_key"] = api_key_secret.get_secret_value()
+            elif ad_token_provider:
+                client_params["azure_ad_token_provider"] = ad_token_provider
+
+            if resolved_base_url := azure_openai_settings.get("base_url"):
+                client_params["base_url"] = str(resolved_base_url)
+            elif resolved_endpoint := azure_openai_settings.get("endpoint"):
+                client_params["azure_endpoint"] = str(resolved_endpoint)
+
+            async_client = AsyncAzureOpenAI(**client_params)
+
+        super().__init__(
+            model_id=chat_deployment_name,
+            assistant_id=assistant_id,
+            assistant_name=assistant_name,
+            assistant_description=assistant_description,
+            thread_id=thread_id,
+            async_client=async_client,  # type: ignore[reportArgumentType]
+            default_headers=default_headers,
+        )
+
+
+# endregion
+
+
+# region AzureOpenAIEmbeddingClient
+
+
+AzureOpenAIEmbeddingOptionsT = TypeVar(
+    "AzureOpenAIEmbeddingOptionsT",
+    bound=TypedDict,  # type: ignore[valid-type]
+    default="OpenAIEmbeddingOptions",
+    covariant=True,
+)
+
+
+@deprecated("AzureOpenAIEmbeddingClient is deprecated. Use OpenAIEmbeddingClient with an AsyncAzureOpenAI client.")
+class AzureOpenAIEmbeddingClient(
+    EmbeddingTelemetryLayer[str, list[float], AzureOpenAIEmbeddingOptionsT],
+    RawOpenAIEmbeddingClient[AzureOpenAIEmbeddingOptionsT],
+    Generic[AzureOpenAIEmbeddingOptionsT],
+):
+    """Deprecated Azure OpenAI embedding client. Use OpenAIEmbeddingClient with AsyncAzureOpenAI instead."""
+
+    OTEL_PROVIDER_NAME: ClassVar[str] = "azure.ai.openai"
+
+    def __init__(
+        self,
+        *,
+        api_key: str | None = None,
+        deployment_name: str | None = None,
+        endpoint: str | None = None,
+        base_url: str | None = None,
+        api_version: str | None = None,
+        token_endpoint: str | None = None,
+        credential: AzureCredentialTypes | AzureTokenProvider | None = None,
+        default_headers: Mapping[str, str] | None = None,
+        async_client: AsyncAzureOpenAI | None = None,
+        otel_provider_name: str | None = None,
+        env_file_path: str | None = None,
+        env_file_encoding: str | None = None,
+    ) -> None:
+        """Initialize an Azure OpenAI embedding client.
+
+        Keyword Args:
+            api_key: The API key.
+            deployment_name: The deployment name.
+            endpoint: The deployment endpoint.
+            base_url: The deployment base URL.
+            api_version: The deployment API version.
+            token_endpoint: The token endpoint to request an Azure token.
+            credential: Azure credential or token provider for authentication.
+            default_headers: Default headers for HTTP requests.
+            async_client: An existing client to use.
+            otel_provider_name: Override the OpenTelemetry provider name.
+            env_file_path: Path to .env file for settings.
+            env_file_encoding: Encoding for .env file.
+        """
+        azure_openai_settings = load_settings(
+            AzureOpenAISettings,
+            env_prefix="AZURE_OPENAI_",
+            api_key=api_key,
+            base_url=base_url,
+            endpoint=endpoint,
+            embedding_deployment_name=deployment_name,
+            api_version=api_version,
+            env_file_path=env_file_path,
+            env_file_encoding=env_file_encoding,
+            token_endpoint=token_endpoint,
+        )
+        _apply_azure_defaults(azure_openai_settings)
+
+        embedding_deployment_name = azure_openai_settings.get("embedding_deployment_name")
+        if not embedding_deployment_name:
+            raise ValueError(
+                "Azure OpenAI embedding deployment name is required. Set via 'deployment_name' parameter "
+                "or 'AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME' environment variable."
+            )
+
+        if not async_client:
+            # Create the Azure OpenAI client directly
+            merged_headers = dict(copy(default_headers)) if default_headers else {}
+            if APP_INFO:
+                merged_headers.update(APP_INFO)
+                merged_headers = prepend_agent_framework_to_user_agent(merged_headers)
+
+            api_key_secret = azure_openai_settings.get("api_key")
+            ad_token_provider = None
+            if not api_key_secret and credential:
+                ad_token_provider = resolve_credential_to_token_provider(
+                    credential, azure_openai_settings.get("token_endpoint")
+                )
+
+            if not api_key_secret and not ad_token_provider:
+                raise ValueError("Please provide either api_key, credential, or a client.")
+
+            endpoint_value = azure_openai_settings.get("endpoint")
+            base_url_value = azure_openai_settings.get("base_url")
+            if not endpoint_value and not base_url_value:
+                raise ValueError("Please provide an endpoint or a base_url")
+
+            client_args: dict[str, Any] = {"default_headers": merged_headers}
+            if resolved_api_version := azure_openai_settings.get("api_version"):
+                client_args["api_version"] = resolved_api_version
+            if ad_token_provider:
+                client_args["azure_ad_token_provider"] = ad_token_provider
+            if api_key_secret:
+                client_args["api_key"] = api_key_secret.get_secret_value()
+            if base_url_value:
+                client_args["base_url"] = str(base_url_value)
+            if endpoint_value and not base_url_value:
+                client_args["azure_endpoint"] = str(endpoint_value)
+            if embedding_deployment_name:
+                client_args["azure_deployment"] = embedding_deployment_name
+
+            async_client = AsyncAzureOpenAI(**client_args)
+
+        # Store Azure-specific attributes for serialization
+        self.endpoint = str(azure_openai_settings.get("endpoint") or "")
+        self.api_version = azure_openai_settings.get("api_version") or ""
+        self.deployment_name = embedding_deployment_name
+
+        super().__init__(
+            async_client=async_client,
+            model=embedding_deployment_name,
+            default_headers=default_headers,
+        )
+        if otel_provider_name is not None:
+            self.OTEL_PROVIDER_NAME = otel_provider_name  # type: ignore[misc]
+
+
+# endregion
@@ -6,11 +6,10 @@ import logging
 from collections.abc import Awaitable, Callable
 from typing import Union

+from agent_framework.exceptions import ChatClientInvalidAuthException
 from azure.core.credentials import TokenCredential
 from azure.core.credentials_async import AsyncTokenCredential

-from ..exceptions import ChatClientInvalidAuthException
-
 logger: logging.Logger = logging.getLogger(__name__)

 AzureTokenProvider = Callable[[], Union[str, Awaitable[str]]]
@@ -18,7 +18,6 @@ from agent_framework import (
 from agent_framework._mcp import MCPTool
 from agent_framework._settings import load_settings
 from agent_framework._tools import ToolTypes
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
 from azure.ai.projects.aio import AIProjectClient
 from azure.ai.projects.models import (
    AgentVersionDetails,
@@ -29,13 +28,15 @@ from azure.ai.projects.models import (
    FunctionTool as AzureFunctionTool,
 )

-from ._client import AzureAIClient, AzureAIProjectAgentOptions
+from ._client import AzureAIClient, AzureAIProjectAgentOptions  # pyright: ignore[reportDeprecated]
+from ._entra_id_authentication import AzureCredentialTypes
 from ._shared import AzureAISettings, create_text_format_config, from_azure_ai_tools, to_azure_ai_tools

 if sys.version_info >= (3, 13):
    from typing import TypeVar  # type: ignore # pragma: no cover
+    from warnings import deprecated  # type: ignore # pragma: no cover
 else:
-    from typing_extensions import TypeVar  # type: ignore # pragma: no cover
+    from typing_extensions import TypeVar, deprecated  # type: ignore # pragma: no cover
 if sys.version_info >= (3, 11):
    from typing import Self, TypedDict  # type: ignore # pragma: no cover
 else:
@@ -55,11 +56,12 @@ OptionsCoT = TypeVar(
 )


+@deprecated("AzureAIProjectAgentProvider is deprecated. Use FoundryAgent instead.")
 class AzureAIProjectAgentProvider(Generic[OptionsCoT]):
-    """Provider for Azure AI Agent Service (Responses API).
+    """Deprecated provider for Azure AI Agent Service (Responses API).

-    This provider allows you to create, retrieve, and manage Azure AI agents
-    using the AIProjectClient from the Azure AI Projects SDK.
+    This provider is deprecated. Use ``FoundryAgent`` instead to connect to
+    pre-configured agents in Foundry.

    Examples:
        Using with explicit AIProjectClient:
@@ -200,7 +202,7 @@ class AzureAIProjectAgentProvider(Generic[OptionsCoT]):
            )

        # Extract options from default_options if present
-        opts = dict(default_options) if default_options else {}
+        opts: dict[str, Any] = dict(default_options) if default_options else {}
        response_format = opts.get("response_format")
        rai_config = opts.get("rai_config")
        reasoning = opts.get("reasoning")
@@ -384,7 +386,7 @@ class AzureAIProjectAgentProvider(Generic[OptionsCoT]):
        if not isinstance(details.definition, PromptAgentDefinition):
            raise ValueError("Agent definition must be PromptAgentDefinition to get a Agent.")

-        client = AzureAIClient(
+        client = AzureAIClient(  # pyright: ignore[reportDeprecated]
            project_client=self._project_client,
            agent_name=details.name,
            agent_version=details.version,
@@ -24,8 +24,11 @@ classifiers = [
 ]
 dependencies = [
    "agent-framework-core>=1.0.0rc5",
+    "agent-framework-openai>=1.0.0rc5",
+    "azure-ai-projects>=2.0.0,<3.0",
    "azure-ai-agents>=1.2.0b5,<1.2.0b6",
    "azure-ai-inference>=1.0.0b9,<1.0.0b10",
+    "azure-identity>=1,<2",
    "aiohttp>=3.7.0,<4",
 ]

@@ -1,9 +1,8 @@
 # Copyright (c) Microsoft. All rights reserved.
 from typing import Any

-from pytest import fixture
-
 from agent_framework import Message
+from pytest import fixture


 # region: Connector Settings fixtures
@@ -1,31 +1,16 @@
 # Copyright (c) Microsoft. All rights reserved.

-import os
 from typing import Annotated
 from unittest.mock import AsyncMock, MagicMock, patch

 import pytest
-from azure.identity import AzureCliCredential
-from pydantic import Field
-
 from agent_framework import (
-    Agent,
-    AgentResponse,
-    AgentResponseUpdate,
-    AgentSession,
-    ChatResponse,
-    ChatResponseUpdate,
-    Message,
    SupportsChatGetResponse,
    tool,
 )
 from agent_framework._settings import SecretString
 from agent_framework.azure import AzureOpenAIAssistantsClient
-
-skip_if_azure_integration_tests_disabled = pytest.mark.skipif(
-    os.getenv("AZURE_OPENAI_ENDPOINT", "") in ("", "https://test-endpoint.com"),
-    reason="No real AZURE_OPENAI_ENDPOINT provided; skipping integration tests.",
-)
+from pydantic import Field


 def create_test_azure_assistants_client(
@@ -87,7 +72,7 @@ def test_azure_assistants_client_init_with_client(mock_async_azure_openai: Magic
    )

    assert client.client is mock_async_azure_openai
-    assert client.model_id == "test_chat_deployment"
+    assert client.model == "test_chat_deployment"
    assert client.assistant_id == "existing-assistant-id"
    assert client.thread_id == "test-thread-id"
    assert not client._should_delete_assistant  # type: ignore
@@ -108,7 +93,7 @@ def test_azure_assistants_client_init_auto_create_client(
    )

    assert client.client is mock_async_azure_openai
-    assert client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
+    assert client.model == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
    assert client.assistant_id is None
    assert client.assistant_name == "TestAssistant"
    assert not client._should_delete_assistant  # type: ignore
@@ -139,7 +124,7 @@ def test_azure_assistants_client_init_with_default_headers(azure_openai_unit_tes
        default_headers=default_headers,
    )

-    assert client.model_id == "test_chat_deployment"
+    assert client.model == "test_chat_deployment"
    assert isinstance(client, SupportsChatGetResponse)

    # Assert that the default header we added is present in the client's default headers
@@ -235,7 +220,7 @@ def test_azure_assistants_client_serialize(azure_openai_unit_test_env: dict[str,

    dumped_settings = client.to_dict()

-    assert dumped_settings["model_id"] == "test_chat_deployment"
+    assert dumped_settings["model"] == "test_chat_deployment"
    assert dumped_settings["assistant_id"] == "test-assistant-id"
    assert dumped_settings["assistant_name"] == "TestAssistant"
    assert dumped_settings["thread_id"] == "test-thread-id"
@@ -256,319 +241,18 @@ def get_weather(
    return f"The weather in {location} is sunny with a high of 25°C."


-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_get_response() -> None:
-    """Test Azure Assistants Client response."""
-    async with AzureOpenAIAssistantsClient(credential=AzureCliCredential()) as azure_assistants_client:
-        assert isinstance(azure_assistants_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(
-            Message(
-                role="user",
-                text="The weather in Seattle is currently sunny with a high of 25°C. "
-                "It's a beautiful day for outdoor activities.",
-            )
-        )
-        messages.append(Message(role="user", text="What's the weather like today?"))
-
-        # Test that the client can be used to get a response
-        response = await azure_assistants_client.get_response(messages=messages)
-
-        assert response is not None
-        assert isinstance(response, ChatResponse)
-        assert any(word in response.text.lower() for word in ["sunny", "25", "weather", "seattle"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_get_response_tools() -> None:
-    """Test Azure Assistants Client response with tools."""
-    async with AzureOpenAIAssistantsClient(credential=AzureCliCredential()) as azure_assistants_client:
-        assert isinstance(azure_assistants_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(Message(role="user", text="What's the weather like in Seattle?"))
-
-        # Test that the client can be used to get a response
-        response = await azure_assistants_client.get_response(
-            messages=messages,
-            options={"tools": [get_weather], "tool_choice": "auto"},
-        )
-
-        assert response is not None
-        assert isinstance(response, ChatResponse)
-        assert any(word in response.text.lower() for word in ["sunny", "25", "weather"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_streaming() -> None:
-    """Test Azure Assistants Client streaming response."""
-    async with AzureOpenAIAssistantsClient(credential=AzureCliCredential()) as azure_assistants_client:
-        assert isinstance(azure_assistants_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(
-            Message(
-                role="user",
-                text="The weather in Seattle is currently sunny with a high of 25°C. "
-                "It's a beautiful day for outdoor activities.",
-            )
-        )
-        messages.append(Message(role="user", text="What's the weather like today?"))
-
-        # Test that the client can be used to get a response
-        response = azure_assistants_client.get_response(messages=messages, stream=True)
-
-        full_message: str = ""
-        async for chunk in response:
-            assert chunk is not None
-            assert isinstance(chunk, ChatResponseUpdate)
-            for content in chunk.contents:
-                if content.type == "text" and content.text:
-                    full_message += content.text
-
-        assert any(word in full_message.lower() for word in ["sunny", "25", "weather", "seattle"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_streaming_tools() -> None:
-    """Test Azure Assistants Client streaming response with tools."""
-    async with AzureOpenAIAssistantsClient(credential=AzureCliCredential()) as azure_assistants_client:
-        assert isinstance(azure_assistants_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(Message(role="user", text="What's the weather like in Seattle?"))
-
-        # Test that the client can be used to get a response
-        response = azure_assistants_client.get_response(
-            messages=messages,
-            options={"tools": [get_weather], "tool_choice": "auto"},
-            stream=True,
-        )
-        full_message: str = ""
-        async for chunk in response:
-            assert chunk is not None
-            assert isinstance(chunk, ChatResponseUpdate)
-            for content in chunk.contents:
-                if content.type == "text" and content.text:
-                    full_message += content.text
-
-        assert any(word in full_message.lower() for word in ["sunny", "25", "weather"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_with_existing_assistant() -> None:
-    """Test Azure Assistants Client with existing assistant ID."""
-    # First create an assistant to use in the test
-    async with AzureOpenAIAssistantsClient(credential=AzureCliCredential()) as temp_client:
-        # Get the assistant ID by triggering assistant creation
-        messages = [Message(role="user", text="Hello")]
-        await temp_client.get_response(messages=messages)
-        assistant_id = temp_client.assistant_id
-
-        # Now test using the existing assistant
-        async with AzureOpenAIAssistantsClient(
-            assistant_id=assistant_id, credential=AzureCliCredential()
-        ) as azure_assistants_client:
-            assert isinstance(azure_assistants_client, SupportsChatGetResponse)
-            assert azure_assistants_client.assistant_id == assistant_id
-
-            messages = [Message(role="user", text="What can you do?")]
-
-            # Test that the client can be used to get a response
-            response = await azure_assistants_client.get_response(messages=messages)
-
-            assert response is not None
-            assert isinstance(response, ChatResponse)
-            assert len(response.text) > 0
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_agent_basic_run():
-    """Test Agent basic run functionality with AzureOpenAIAssistantsClient."""
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-    ) as agent:
-        # Run a simple query
-        response = await agent.run("Hello! Please respond with 'Hello World' exactly.")
-
-        # Validate response
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        assert len(response.text) > 0
-        assert "Hello World" in response.text
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_agent_basic_run_streaming():
-    """Test Agent basic streaming functionality with AzureOpenAIAssistantsClient."""
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-    ) as agent:
-        # Run streaming query
-        full_message: str = ""
-        async for chunk in agent.run("Please respond with exactly: 'This is a streaming response test.'", stream=True):
-            assert chunk is not None
-            assert isinstance(chunk, AgentResponseUpdate)
-            if chunk.text:
-                full_message += chunk.text
-
-        # Validate streaming response
-        assert len(full_message) > 0
-        assert "streaming response test" in full_message.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_agent_session_persistence():
-    """Test Agent session persistence across runs with AzureOpenAIAssistantsClient."""
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant with good memory.",
-    ) as agent:
-        # Create a new session that will be reused
-        session = agent.create_session()
-
-        # First message - establish context
-        first_response = await agent.run(
-            "Remember this number: 42. What number did I just tell you to remember?", session=session
-        )
-        assert isinstance(first_response, AgentResponse)
-        assert "42" in first_response.text
-
-        # Second message - test conversation memory
-        second_response = await agent.run(
-            "What number did I tell you to remember in my previous message?", session=session
-        )
-        assert isinstance(second_response, AgentResponse)
-        assert "42" in second_response.text
-
-        # Verify session has been populated with conversation ID
-        assert session.service_session_id is not None
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_agent_existing_session_id():
-    """Test Agent with existing session ID to continue conversations across agent instances."""
-    # First, create a conversation and capture the session ID
-    existing_session_id = None
-
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-        instructions="You are a helpful weather agent.",
-        tools=[get_weather],
-    ) as agent:
-        # Start a conversation and get the session ID
-        session = agent.create_session()
-        response1 = await agent.run("What's the weather in Paris?", session=session)
-
-        # Validate first response
-        assert isinstance(response1, AgentResponse)
-        assert response1.text is not None
-        assert any(word in response1.text.lower() for word in ["weather", "paris"])
-
-        # The session ID is set after the first response
-        existing_session_id = session.service_session_id
-        assert existing_session_id is not None
-
-    # Now continue with the same session ID in a new agent instance
-
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(thread_id=existing_session_id, credential=AzureCliCredential()),
-        instructions="You are a helpful weather agent.",
-        tools=[get_weather],
-    ) as agent:
-        # Create a session with the existing ID
-        session = AgentSession(service_session_id=existing_session_id)
-
-        # Ask about the previous conversation
-        response2 = await agent.run("What was the last city I asked about?", session=session)
-
-        # Validate that the agent remembers the previous conversation
-        assert isinstance(response2, AgentResponse)
-        assert response2.text is not None
-        # Should reference Paris from the previous conversation
-        assert "paris" in response2.text.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_agent_code_interpreter():
-    """Test Agent with code interpreter through AzureOpenAIAssistantsClient."""
-
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant that can write and execute Python code.",
-        tools=[AzureOpenAIAssistantsClient.get_code_interpreter_tool()],
-    ) as agent:
-        # Request code execution
-        response = await agent.run("Write Python code to calculate the factorial of 5 and show the result.")
-
-        # Validate response
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        # Factorial of 5 is 120
-        assert "120" in response.text or "factorial" in response.text.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_integration_tests_disabled
-async def test_azure_assistants_client_agent_level_tool_persistence():
-    """Test that agent-level tools persist across multiple runs with Azure Assistants Client."""
-
-    async with Agent(
-        client=AzureOpenAIAssistantsClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant that uses available tools.",
-        tools=[get_weather],  # Agent-level tool
-    ) as agent:
-        # First run - agent-level tool should be available
-        first_response = await agent.run("What's the weather like in Chicago?")
-
-        assert isinstance(first_response, AgentResponse)
-        assert first_response.text is not None
-        # Should use the agent-level weather tool
-        assert any(term in first_response.text.lower() for term in ["chicago", "sunny", "72"])
-
-        # Second run - agent-level tool should still be available (persistence test)
-        second_response = await agent.run("What's the weather in Miami?")
-
-        assert isinstance(second_response, AgentResponse)
-        assert second_response.text is not None
-        # Should use the agent-level weather tool again
-        assert any(term in second_response.text.lower() for term in ["miami", "sunny", "72"])
-
-
 def test_azure_assistants_client_entra_id_authentication() -> None:
    """Test credential authentication path with sync credential."""
    mock_credential = MagicMock()
    mock_provider = MagicMock(return_value="token-string")

    with (
-        patch("agent_framework.azure._assistants_client.load_settings") as mock_load_settings,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.load_settings") as mock_load_settings,
        patch(
-            "agent_framework.azure._assistants_client.resolve_credential_to_token_provider",
+            "agent_framework_azure_ai._deprecated_azure_openai.resolve_credential_to_token_provider",
            return_value=mock_provider,
        ) as mock_resolve,
-        patch("agent_framework.azure._assistants_client.AsyncAzureOpenAI") as mock_azure_client,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.AsyncAzureOpenAI") as mock_azure_client,
        patch("agent_framework.openai.OpenAIAssistantsClient.__init__", return_value=None),
    ):
        mock_load_settings.return_value = {
@@ -602,7 +286,7 @@ def test_azure_assistants_client_entra_id_authentication() -> None:

 def test_azure_assistants_client_no_authentication_error() -> None:
    """Test authentication validation error when no auth provided."""
-    with patch("agent_framework.azure._assistants_client.load_settings") as mock_load_settings:
+    with patch("agent_framework_azure_ai._deprecated_azure_openai.load_settings") as mock_load_settings:
        mock_load_settings.return_value = {
            "chat_deployment_name": "test-deployment",
            "responses_deployment_name": None,
@@ -627,12 +311,12 @@ def test_azure_assistants_client_callable_credential() -> None:
    mock_provider = MagicMock(return_value="my-token")

    with (
-        patch("agent_framework.azure._assistants_client.load_settings") as mock_load_settings,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.load_settings") as mock_load_settings,
        patch(
-            "agent_framework.azure._assistants_client.resolve_credential_to_token_provider",
+            "agent_framework_azure_ai._deprecated_azure_openai.resolve_credential_to_token_provider",
            return_value=mock_provider,
        ),
-        patch("agent_framework.azure._assistants_client.AsyncAzureOpenAI") as mock_azure_client,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.AsyncAzureOpenAI") as mock_azure_client,
        patch("agent_framework.openai.OpenAIAssistantsClient.__init__", return_value=None),
    ):
        mock_load_settings.return_value = {
@@ -664,8 +348,8 @@ def test_azure_assistants_client_callable_credential() -> None:
 def test_azure_assistants_client_base_url_configuration() -> None:
    """Test base_url client parameter path."""
    with (
-        patch("agent_framework.azure._assistants_client.load_settings") as mock_load_settings,
-        patch("agent_framework.azure._assistants_client.AsyncAzureOpenAI") as mock_azure_client,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.load_settings") as mock_load_settings,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.AsyncAzureOpenAI") as mock_azure_client,
        patch("agent_framework.openai.OpenAIAssistantsClient.__init__", return_value=None),
    ):
        mock_load_settings.return_value = {
@@ -695,8 +379,8 @@ def test_azure_assistants_client_base_url_configuration() -> None:
 def test_azure_assistants_client_azure_endpoint_configuration() -> None:
    """Test azure_endpoint client parameter path."""
    with (
-        patch("agent_framework.azure._assistants_client.load_settings") as mock_load_settings,
-        patch("agent_framework.azure._assistants_client.AsyncAzureOpenAI") as mock_azure_client,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.load_settings") as mock_load_settings,
+        patch("agent_framework_azure_ai._deprecated_azure_openai.AsyncAzureOpenAI") as mock_azure_client,
        patch("agent_framework.openai.OpenAIAssistantsClient.__init__", return_value=None),
    ):
        mock_load_settings.return_value = {
@@ -6,16 +6,6 @@ from unittest.mock import AsyncMock, MagicMock, patch

 import openai
 import pytest
-from azure.identity import AzureCliCredential
-from httpx import Request, Response
-from openai import AsyncAzureOpenAI, AsyncStream
-from openai.resources.chat.completions import AsyncCompletions as AsyncChatCompletions
-from openai.types.chat import ChatCompletion, ChatCompletionChunk
-from openai.types.chat.chat_completion import Choice
-from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
-from openai.types.chat.chat_completion_chunk import ChoiceDelta as ChunkChoiceDelta
-from openai.types.chat.chat_completion_message import ChatCompletionMessage
-
 from agent_framework import (
    Agent,
    AgentResponse,
@@ -29,10 +19,19 @@ from agent_framework import (
 from agent_framework._telemetry import USER_AGENT_KEY
 from agent_framework.azure import AzureOpenAIChatClient
 from agent_framework.exceptions import ChatClientException
-from agent_framework.openai import (
+from agent_framework_openai import (
    ContentFilterResultSeverity,
    OpenAIContentFilterException,
 )
+from azure.identity import AzureCliCredential
+from httpx import Request, Response
+from openai import AsyncAzureOpenAI, AsyncStream
+from openai.resources.chat.completions import AsyncCompletions as AsyncChatCompletions
+from openai.types.chat import ChatCompletion, ChatCompletionChunk
+from openai.types.chat.chat_completion import Choice
+from openai.types.chat.chat_completion_chunk import Choice as ChunkChoice
+from openai.types.chat.chat_completion_chunk import ChoiceDelta as ChunkChoiceDelta
+from openai.types.chat.chat_completion_message import ChatCompletionMessage

 # region Service Setup

@@ -48,7 +47,7 @@ def test_init(azure_openai_unit_test_env: dict[str, str]) -> None:

    assert azure_chat_client.client is not None
    assert isinstance(azure_chat_client.client, AsyncAzureOpenAI)
-    assert azure_chat_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
+    assert azure_chat_client.model == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
    assert isinstance(azure_chat_client, SupportsChatGetResponse)


@@ -71,7 +70,7 @@ def test_init_base_url(azure_openai_unit_test_env: dict[str, str]) -> None:

    assert azure_chat_client.client is not None
    assert isinstance(azure_chat_client.client, AsyncAzureOpenAI)
-    assert azure_chat_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
+    assert azure_chat_client.model == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
    assert isinstance(azure_chat_client, SupportsChatGetResponse)
    for key, value in default_headers.items():
        assert key in azure_chat_client.client.default_headers
@@ -84,7 +83,7 @@ def test_init_endpoint(azure_openai_unit_test_env: dict[str, str]) -> None:

    assert azure_chat_client.client is not None
    assert isinstance(azure_chat_client.client, AsyncAzureOpenAI)
-    assert azure_chat_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
+    assert azure_chat_client.model == azure_openai_unit_test_env["AZURE_OPENAI_CHAT_DEPLOYMENT_NAME"]
    assert isinstance(azure_chat_client, SupportsChatGetResponse)


@@ -131,7 +130,7 @@ def test_serialize(azure_openai_unit_test_env: dict[str, str]) -> None:

    azure_chat_client = AzureOpenAIChatClient.from_dict(settings)
    dumped_settings = azure_chat_client.to_dict()
-    assert dumped_settings["model_id"] == settings["deployment_name"]
+    assert dumped_settings["model"] == settings["deployment_name"]
    assert str(settings["endpoint"]) in str(dumped_settings["endpoint"])
    assert str(settings["deployment_name"]) == str(dumped_settings["deployment_name"])
    assert settings["api_version"] == dumped_settings["api_version"]
@@ -6,13 +6,12 @@ import os
 from unittest.mock import AsyncMock, MagicMock

 import pytest
+from agent_framework.azure import AzureOpenAIEmbeddingClient
+from agent_framework_openai import OpenAIEmbeddingOptions
 from openai.types import CreateEmbeddingResponse
 from openai.types import Embedding as OpenAIEmbedding
 from openai.types.create_embedding_response import Usage

-from agent_framework.azure import AzureOpenAIEmbeddingClient
-from agent_framework.openai import OpenAIEmbeddingOptions
-

 def _make_openai_response(
    embeddings: list[list[float]],
@@ -49,7 +48,7 @@ def test_azure_construction_with_deployment_name(azure_embedding_unit_test_env:
        api_key="test-key",
        endpoint="https://test.openai.azure.com/",
    )
-    assert client.model_id == "text-embedding-3-small"
+    assert client.model == "text-embedding-3-small"


 def test_azure_construction_with_existing_client(azure_embedding_unit_test_env: None) -> None:
@@ -58,7 +57,7 @@ def test_azure_construction_with_existing_client(azure_embedding_unit_test_env:
        deployment_name="my-deployment",
        async_client=mock_client,
    )
-    assert client.model_id == "my-deployment"
+    assert client.model == "my-deployment"
    assert client.client is mock_client


@@ -8,10 +8,6 @@ from typing import Annotated, Any
 from unittest.mock import MagicMock

 import pytest
-from azure.identity import AzureCliCredential
-from pydantic import BaseModel
-from pytest import param
-
 from agent_framework import (
    Agent,
    AgentResponse,
@@ -22,6 +18,9 @@ from agent_framework import (
    tool,
 )
 from agent_framework.azure import AzureOpenAIResponsesClient
+from azure.identity import AzureCliCredential
+from pydantic import BaseModel
+from pytest import param

 skip_if_azure_integration_tests_disabled = pytest.mark.skipif(
    os.getenv("AZURE_OPENAI_ENDPOINT", "") in ("", "https://test-endpoint.com"),
@@ -75,7 +74,7 @@ def test_init(azure_openai_unit_test_env: dict[str, str]) -> None:
    # Test successful initialization
    azure_responses_client = AzureOpenAIResponsesClient(credential=AzureCliCredential())

-    assert azure_responses_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]
+    assert azure_responses_client.model == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]
    assert isinstance(azure_responses_client, SupportsChatGetResponse)


@@ -90,7 +89,7 @@ def test_init_model_id_constructor(azure_openai_unit_test_env: dict[str, str]) -
    model_id = "test_model_id"
    azure_responses_client = AzureOpenAIResponsesClient(deployment_name=model_id)

-    assert azure_responses_client.model_id == model_id
+    assert azure_responses_client.model == model_id
    assert isinstance(azure_responses_client, SupportsChatGetResponse)


@@ -98,7 +97,7 @@ def test_init_model_id_kwarg(azure_openai_unit_test_env: dict[str, str]) -> None
    """Test that model_id kwarg correctly sets the deployment name (issue #4299)."""
    azure_responses_client = AzureOpenAIResponsesClient(model_id="gpt-4o")

-    assert azure_responses_client.model_id == "gpt-4o"
+    assert azure_responses_client.model == "gpt-4o"
    assert isinstance(azure_responses_client, SupportsChatGetResponse)


@@ -108,7 +107,7 @@ def test_init_model_id_kwarg_does_not_override_deployment_name(
    """Test that deployment_name takes precedence over model_id kwarg (issue #4299)."""
    azure_responses_client = AzureOpenAIResponsesClient(deployment_name="my-deployment", model_id="gpt-4o")

-    assert azure_responses_client.model_id == "my-deployment"
+    assert azure_responses_client.model == "my-deployment"
    assert isinstance(azure_responses_client, SupportsChatGetResponse)


@@ -116,7 +115,7 @@ def test_init_model_id_kwarg_none(azure_openai_unit_test_env: dict[str, str]) ->
    """Test that model_id=None does not override the env-var deployment name."""
    azure_responses_client = AzureOpenAIResponsesClient(model_id=None)

-    assert azure_responses_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]
+    assert azure_responses_client.model == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]


 def test_init_with_default_header(azure_openai_unit_test_env: dict[str, str]) -> None:
@@ -127,7 +126,7 @@ def test_init_with_default_header(azure_openai_unit_test_env: dict[str, str]) ->
        default_headers=default_headers,
    )

-    assert azure_responses_client.model_id == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]
+    assert azure_responses_client.model == azure_openai_unit_test_env["AZURE_OPENAI_RESPONSES_DEPLOYMENT_NAME"]
    assert isinstance(azure_responses_client, SupportsChatGetResponse)

    # Assert that the default header we added is present in the client's default headers
@@ -156,7 +155,7 @@ def test_init_with_project_client(azure_openai_unit_test_env: dict[str, str]) ->
    mock_project_client.get_openai_client.return_value = mock_openai_client

    with patch(
-        "agent_framework.azure._responses_client.AzureOpenAIResponsesClient._create_client_from_project",
+        "agent_framework_azure_ai._deprecated_azure_openai.AzureOpenAIResponsesClient._create_client_from_project",
        return_value=mock_openai_client,
    ):
        azure_responses_client = AzureOpenAIResponsesClient(
@@ -164,7 +163,7 @@ def test_init_with_project_client(azure_openai_unit_test_env: dict[str, str]) ->
            deployment_name="gpt-4o",
        )

-    assert azure_responses_client.model_id == "gpt-4o"
+    assert azure_responses_client.model == "gpt-4o"
    assert azure_responses_client.client is mock_openai_client
    assert isinstance(azure_responses_client, SupportsChatGetResponse)

@@ -179,7 +178,7 @@ def test_init_with_project_endpoint(azure_openai_unit_test_env: dict[str, str])
    mock_openai_client.default_headers = {}

    with patch(
-        "agent_framework.azure._responses_client.AzureOpenAIResponsesClient._create_client_from_project",
+        "agent_framework_azure_ai._deprecated_azure_openai.AzureOpenAIResponsesClient._create_client_from_project",
        return_value=mock_openai_client,
    ):
        azure_responses_client = AzureOpenAIResponsesClient(
@@ -188,7 +187,7 @@ def test_init_with_project_endpoint(azure_openai_unit_test_env: dict[str, str])
            credential=AzureCliCredential(),
        )

-    assert azure_responses_client.model_id == "gpt-4o"
+    assert azure_responses_client.model == "gpt-4o"
    assert azure_responses_client.client is mock_openai_client
    assert isinstance(azure_responses_client, SupportsChatGetResponse)

@@ -220,7 +219,7 @@ def test_create_client_from_project_with_endpoint() -> None:
    mock_openai_client = MagicMock(spec=AsyncOpenAI)
    mock_credential = MagicMock()

-    with patch("agent_framework.azure._responses_client.AIProjectClient") as MockAIProjectClient:
+    with patch("agent_framework_azure_ai._deprecated_azure_openai.AIProjectClient") as MockAIProjectClient:
        mock_instance = MockAIProjectClient.return_value
        mock_instance.get_openai_client.return_value = mock_openai_client

@@ -668,8 +667,8 @@ async def test_azure_openai_responses_client_tool_rich_content_image() -> None:

 skip_if_azure_ai_integration_tests_disabled = pytest.mark.skipif(
    os.getenv("AZURE_AI_PROJECT_ENDPOINT", "") in ("", "https://test-project.cognitiveservices.azure.com/")
-    or os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME", "") == "",
-    reason="No real AZURE_AI_PROJECT_ENDPOINT or AZURE_AI_MODEL_DEPLOYMENT_NAME provided; skipping integration tests.",
+    or os.getenv("AZURE_AI_MODEL", "") == "",
+    reason="No real AZURE_AI_PROJECT_ENDPOINT or AZURE_AI_MODEL provided; skipping integration tests.",
 )


@@ -695,7 +694,7 @@ async def test_integration_function_call_roundtrip_preserves_fidelity():

    client = AzureOpenAIResponsesClient(
        project_endpoint=os.environ["AZURE_AI_PROJECT_ENDPOINT"],
-        deployment_name=os.environ["AZURE_AI_MODEL_DEPLOYMENT_NAME"],
+        deployment_name=os.environ["AZURE_AI_MODEL"],
        credential=AzureCliCredential(),
    )

@@ -3,13 +3,13 @@
 from unittest.mock import MagicMock, patch

 import pytest
+from agent_framework.exceptions import ChatClientInvalidAuthException
 from azure.core.credentials import TokenCredential
 from azure.core.credentials_async import AsyncTokenCredential

-from agent_framework.azure._entra_id_authentication import (
+from agent_framework_azure_ai._entra_id_authentication import (
    resolve_credential_to_token_provider,
 )
-from agent_framework.exceptions import ChatClientInvalidAuthException

 TOKEN_ENDPOINT = "https://cognitiveservices.azure.com/.default"

@@ -15,7 +15,6 @@ from azure.ai.agents.models import (
 from azure.ai.agents.models import (
    CodeInterpreterToolDefinition,
 )
-from azure.identity.aio import AzureCliCredential
 from pydantic import BaseModel

 from agent_framework_azure_ai import (
@@ -772,82 +771,3 @@ def test_from_azure_ai_agent_tools_unknown_dict() -> None:


 # endregion
-
-# region Integration Tests
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_create_agent() -> None:
-    """Integration test: Create an agent using the provider."""
-    async with (
-        AzureCliCredential() as credential,
-        AzureAIAgentsProvider(credential=credential) as provider,
-    ):
-        agent = await provider.create_agent(
-            name="IntegrationTestAgent",
-            instructions="You are a helpful assistant for testing.",
-        )
-
-        try:
-            assert isinstance(agent, Agent)
-            assert agent.name == "IntegrationTestAgent"
-            assert agent.id is not None
-        finally:
-            # Cleanup: delete the agent
-            if agent.id:
-                await provider._agents_client.delete_agent(agent.id)  # type: ignore
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_get_agent() -> None:
-    """Integration test: Get an existing agent using the provider."""
-    async with (
-        AzureCliCredential() as credential,
-        AzureAIAgentsProvider(credential=credential) as provider,
-    ):
-        # First create an agent
-        created = await provider._agents_client.create_agent(  # type: ignore
-            model=os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME", "gpt-4o"),
-            name="GetAgentTest",
-            instructions="Test agent",
-        )
-
-        try:
-            # Then get it using the provider
-            agent = await provider.get_agent(created.id)
-
-            assert isinstance(agent, Agent)
-            assert agent.id == created.id
-        finally:
-            await provider._agents_client.delete_agent(created.id)  # type: ignore
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_create_and_run() -> None:
-    """Integration test: Create an agent and run a conversation."""
-    async with (
-        AzureCliCredential() as credential,
-        AzureAIAgentsProvider(credential=credential) as provider,
-    ):
-        agent = await provider.create_agent(
-            name="RunTestAgent",
-            instructions="You are a helpful assistant. Always respond with 'Hello!' to any greeting.",
-        )
-
-        try:
-            result = await agent.run("Hi there!")
-
-            assert result is not None
-            assert len(result.messages) > 0
-        finally:
-            if agent.id:
-                await provider._agents_client.delete_agent(agent.id)  # type: ignore
-
-
-# endregion
@@ -1,17 +1,11 @@
 # Copyright (c) Microsoft. All rights reserved.

 import json
-import os
-from pathlib import Path
 from typing import Annotated, Any
 from unittest.mock import AsyncMock, MagicMock, patch

 import pytest
 from agent_framework import (
-    Agent,
-    AgentResponse,
-    AgentResponseUpdate,
-    AgentSession,
    ChatOptions,
    ChatResponse,
    ChatResponseUpdate,
@@ -28,7 +22,6 @@ from azure.ai.agents.models import (
    AgentsNamedToolChoiceType,
    AgentsToolChoiceOptionMode,
    CodeInterpreterToolDefinition,
-    FileInfo,
    MessageDeltaChunk,
    MessageDeltaTextContent,
    MessageDeltaTextFileCitationAnnotation,
@@ -41,19 +34,12 @@ from azure.ai.agents.models import (
    SubmitToolApprovalAction,
    SubmitToolOutputsAction,
    ThreadRun,
-    VectorStore,
 )
 from azure.core.credentials_async import AsyncTokenCredential
-from azure.identity.aio import AzureCliCredential
 from pydantic import BaseModel, Field

 from agent_framework_azure_ai import AzureAIAgentClient, AzureAISettings

-skip_if_azure_ai_integration_tests_disabled = pytest.mark.skipif(
-    os.getenv("AZURE_AI_PROJECT_ENDPOINT", "") in ("", "https://test-project.cognitiveservices.azure.com/"),
-    reason="No real AZURE_AI_PROJECT_ENDPOINT provided; skipping integration tests.",
-)
-

 def create_test_azure_ai_chat_client(
    mock_agents_client: MagicMock,
@@ -102,6 +88,15 @@ def create_test_azure_ai_chat_client(
    return client


+def test_init_emits_updated_deprecation_warning(mock_agents_client: MagicMock) -> None:
+    """Test that construction emits the updated class deprecation warning."""
+    with pytest.deprecated_call(match="V1 Agents Service API and has no direct replacement"):
+        AzureAIAgentClient(
+            agents_client=mock_agents_client,
+            agent_id="test-agent",
+        )
+
+
 def test_azure_ai_settings_init(azure_ai_unit_test_env: dict[str, str]) -> None:
    """Test AzureAISettings initialization."""
    settings = load_settings(AzureAISettings, env_prefix="AZURE_AI_")
@@ -1527,401 +1522,6 @@ def get_weather(
    return f"The weather in {location} is sunny with a high of 25°C."


-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_get_response() -> None:
-    """Test Azure AI Chat Client response."""
-    async with AzureAIAgentClient(credential=AzureCliCredential()) as azure_ai_chat_client:
-        assert isinstance(azure_ai_chat_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(
-            Message(
-                role="user",
-                text="The weather in Seattle is currently sunny with a high of 25°C. "
-                "It's a beautiful day for outdoor activities.",
-            )
-        )
-        messages.append(Message(role="user", text="What's the weather like today?"))
-
-        # Test that the agents_client can be used to get a response
-        response = await azure_ai_chat_client.get_response(messages=messages)
-
-        assert response is not None
-        assert isinstance(response, ChatResponse)
-        assert any(word in response.text.lower() for word in ["sunny", "25"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_get_response_tools() -> None:
-    """Test Azure AI Chat Client response with tools."""
-    async with AzureAIAgentClient(credential=AzureCliCredential()) as azure_ai_chat_client:
-        assert isinstance(azure_ai_chat_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(Message(role="user", text="What's the weather like in Seattle?"))
-
-        # Test that the agents_client can be used to get a response
-        response = await azure_ai_chat_client.get_response(
-            messages=messages,
-            options={"tools": [get_weather], "tool_choice": "auto"},
-        )
-
-        assert response is not None
-        assert isinstance(response, ChatResponse)
-        assert any(word in response.text.lower() for word in ["sunny", "25"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_streaming() -> None:
-    """Test Azure AI Chat Client streaming response."""
-    async with AzureAIAgentClient(credential=AzureCliCredential()) as azure_ai_chat_client:
-        assert isinstance(azure_ai_chat_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(
-            Message(
-                role="user",
-                text="The weather in Seattle is currently sunny with a high of 25°C. "
-                "It's a beautiful day for outdoor activities.",
-            )
-        )
-        messages.append(Message(role="user", text="What's the weather like today?"))
-
-        # Test that the agents_client can be used to get a response
-        response = azure_ai_chat_client.get_response(messages=messages, stream=True)
-
-        full_message: str = ""
-        async for chunk in response:
-            assert chunk is not None
-            assert isinstance(chunk, ChatResponseUpdate)
-            for content in chunk.contents:
-                if content.type == "text" and content.text:
-                    full_message += content.text
-
-        assert any(word in full_message.lower() for word in ["sunny", "25"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_streaming_tools() -> None:
-    """Test Azure AI Chat Client streaming response with tools."""
-    async with AzureAIAgentClient(credential=AzureCliCredential()) as azure_ai_chat_client:
-        assert isinstance(azure_ai_chat_client, SupportsChatGetResponse)
-
-        messages: list[Message] = []
-        messages.append(Message(role="user", text="What's the weather like in Seattle?"))
-
-        # Test that the agents_client can be used to get a response
-        response = azure_ai_chat_client.get_response(
-            messages=messages,
-            stream=True,
-            options={"tools": [get_weather], "tool_choice": "auto"},
-        )
-        full_message: str = ""
-        async for chunk in response:
-            assert chunk is not None
-            assert isinstance(chunk, ChatResponseUpdate)
-            for content in chunk.contents:
-                if content.type == "text" and content.text:
-                    full_message += content.text
-
-        assert any(word in full_message.lower() for word in ["sunny", "25"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_basic_run() -> None:
-    """Test Agent basic run functionality with AzureAIAgentClient."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-    ) as agent:
-        # Run a simple query
-        response = await agent.run("Hello! Please respond with 'Hello World' exactly.")
-
-        # Validate response
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        assert len(response.text) > 0
-        assert "Hello World" in response.text
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_basic_run_streaming() -> None:
-    """Test Agent basic streaming functionality with AzureAIAgentClient."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-    ) as agent:
-        # Run streaming query
-        full_message: str = ""
-        async for chunk in agent.run("Please respond with exactly: 'This is a streaming response test.'", stream=True):
-            assert chunk is not None
-            assert isinstance(chunk, AgentResponseUpdate)
-            if chunk.text:
-                full_message += chunk.text
-
-        # Validate streaming response
-        assert len(full_message) > 0
-        assert "streaming response test" in full_message.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_thread_persistence() -> None:
-    """Test Agent session persistence across runs with AzureAIAgentClient."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant with good memory.",
-    ) as agent:
-        # Create a new session that will be reused
-        session = agent.create_session()
-
-        # First message - establish context
-        first_response = await agent.run(
-            "Remember this number: 42. What number did I just tell you to remember?", session=session
-        )
-        assert isinstance(first_response, AgentResponse)
-        assert "42" in first_response.text
-
-        # Second message - test conversation memory
-        second_response = await agent.run(
-            "What number did I tell you to remember in my previous message?", session=session
-        )
-        assert isinstance(second_response, AgentResponse)
-        assert "42" in second_response.text
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_existing_thread_id() -> None:
-    """Test Agent existing thread ID functionality with AzureAIAgentClient."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant with good memory.",
-    ) as first_agent:
-        # Start a conversation and get the session ID
-        session = first_agent.create_session()
-        first_response = await first_agent.run("My name is Alice. Remember this.", session=session)
-
-        # Validate first response
-        assert isinstance(first_response, AgentResponse)
-        assert first_response.text is not None
-
-        # The thread ID is set after the first response
-        existing_thread_id = session.service_session_id
-        assert existing_thread_id is not None
-
-    # Now continue with the same thread ID in a new agent instance
-    async with Agent(
-        client=AzureAIAgentClient(thread_id=existing_thread_id, credential=AzureCliCredential()),
-        instructions="You are a helpful assistant with good memory.",
-    ) as second_agent:
-        # Create a session with the existing ID
-        session = AgentSession(service_session_id=existing_thread_id)
-
-        # Ask about the previous conversation
-        response2 = await second_agent.run("What is my name?", session=session)
-
-        # Validate that the agent remembers the previous conversation
-        assert isinstance(response2, AgentResponse)
-        assert response2.text is not None
-        # Should reference Alice from the previous conversation
-        assert "alice" in response2.text.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_code_interpreter():
-    """Test Agent with code interpreter through AzureAIAgentClient."""
-
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant that can write and execute Python code.",
-        tools=[AzureAIAgentClient.get_code_interpreter_tool()],
-    ) as agent:
-        # Request code execution
-        response = await agent.run("Write Python code to calculate the factorial of 5 and show the result.")
-
-        # Validate response
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        # Factorial of 5 is 120
-        assert "120" in response.text or "factorial" in response.text.lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_file_search():
-    """Test Agent with file search through AzureAIAgentClient."""
-
-    client = AzureAIAgentClient(credential=AzureCliCredential())
-    file: FileInfo | None = None
-    vector_store: VectorStore | None = None
-
-    try:
-        # 1. Read and upload the test file to the Azure AI agent service
-        test_file_path = Path(__file__).parent / "resources" / "employees.pdf"
-        file = await client.agents_client.files.upload_and_poll(file_path=str(test_file_path), purpose="assistants")
-        vector_store = await client.agents_client.vector_stores.create_and_poll(
-            file_ids=[file.id], name="test_employees_vectorstore"
-        )
-
-        # 2. Create file search tool with uploaded resources
-        file_search_tool = AzureAIAgentClient.get_file_search_tool(vector_store_ids=[vector_store.id])
-
-        async with Agent(
-            client=client,
-            instructions="You are a helpful assistant that can search through uploaded employee files.",
-            tools=[file_search_tool],
-        ) as agent:
-            # 3. Test file search functionality
-            response = await agent.run("Who is the youngest employee in the files?")
-
-            # Validate response
-            assert isinstance(response, AgentResponse)
-            assert response.text is not None
-            # Should find information about Alice Johnson (age 24) being the youngest
-            assert any(term in response.text.lower() for term in ["alice", "johnson", "24"])
-
-    finally:
-        # 4. Cleanup: Delete the vector store and file
-        try:
-            if vector_store:
-                await client.agents_client.vector_stores.delete(vector_store.id)
-            if file:
-                await client.agents_client.files.delete(file.id)
-        except Exception:
-            # Ignore cleanup errors to avoid masking the actual test failure
-            pass
-        finally:
-            await client.close()
-
-
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_hosted_mcp_tool() -> None:
-    """Integration test for MCP tool with Azure AI Agent using Microsoft Learn MCP."""
-
-    mcp_tool = AzureAIAgentClient.get_mcp_tool(
-        name="Microsoft Learn MCP",
-        url="https://learn.microsoft.com/api/mcp",
-        description="A Microsoft Learn MCP server for documentation questions",
-        approval_mode="never_require",
-    )
-
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant that can help with microsoft documentation questions.",
-        tools=[mcp_tool],
-    ) as agent:
-        response = await agent.run(
-            "How to create an Azure storage account using az cli?",
-            options={"max_tokens": 200},
-        )
-
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        assert len(response.text) > 0
-
-        # With never_require approval mode, there should be no approval requests
-        assert len(response.user_input_requests) == 0, (
-            f"Expected no approval requests with never_require mode, but got {len(response.user_input_requests)}"
-        )
-
-        # Should contain Azure-related content since it's asking about Azure CLI
-        assert any(term in response.text.lower() for term in ["azure", "storage", "account", "cli"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_level_tool_persistence():
-    """Test that agent-level tools persist across multiple runs with AzureAIAgentClient."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant that uses available tools.",
-        tools=[get_weather],
-    ) as agent:
-        # First run - agent-level tool should be available
-        first_response = await agent.run("What's the weather like in Chicago?")
-
-        assert isinstance(first_response, AgentResponse)
-        assert first_response.text is not None
-        # Should use the agent-level weather tool
-        assert any(term in first_response.text.lower() for term in ["chicago", "sunny", "25"])
-
-        # Second run - agent-level tool should still be available (persistence test)
-        second_response = await agent.run("What's the weather in Miami?")
-
-        assert isinstance(second_response, AgentResponse)
-        assert second_response.text is not None
-        # Should use the agent-level weather tool again
-        assert any(term in second_response.text.lower() for term in ["miami", "sunny", "25"])
-
-
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_chat_options_run_level() -> None:
-    """Test ChatOptions parameter coverage at run level."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant.",
-    ) as agent:
-        response = await agent.run(
-            "Provide a brief, helpful response.",
-            tools=[get_weather],
-            options={
-                "max_tokens": 100,
-                "temperature": 0.7,
-                "top_p": 0.9,
-                "tool_choice": "auto",
-                "metadata": {"test": "value"},
-            },
-        )
-
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        assert len(response.text) > 0
-
-
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_azure_ai_chat_client_agent_chat_options_agent_level() -> None:
-    """Test ChatOptions parameter coverage agent level."""
-    async with Agent(
-        client=AzureAIAgentClient(credential=AzureCliCredential()),
-        instructions="You are a helpful assistant.",
-        tools=[get_weather],
-        default_options={
-            "max_tokens": 100,
-            "temperature": 0.7,
-            "top_p": 0.9,
-            "tool_choice": "auto",
-            "metadata": {"test": "value"},
-        },
-    ) as agent:
-        response = await agent.run(
-            "Provide a brief, helpful response.",
-        )
-
-        assert isinstance(response, AgentResponse)
-        assert response.text is not None
-        assert len(response.text) > 0
-
-
 async def test_azure_ai_chat_client_cleanup_agent_when_enabled_and_created(
    mock_agents_client: MagicMock,
 ) -> None:
@@ -11,8 +11,6 @@ from uuid import uuid4

 import pytest
 from agent_framework import (
-    Agent,
-    AgentResponse,
    Annotation,
    ChatOptions,
    ChatResponse,
@@ -24,7 +22,7 @@ from agent_framework import (
    tool,
 )
 from agent_framework._settings import load_settings
-from agent_framework.openai._responses_client import RawOpenAIResponsesClient
+from agent_framework_openai._chat_client import RawOpenAIChatClient
 from azure.ai.projects.aio import AIProjectClient
 from azure.ai.projects.models import (
    ApproximateLocation,
@@ -41,17 +39,11 @@ from azure.identity.aio import AzureCliCredential
 from openai.types.responses.parsed_response import ParsedResponse
 from openai.types.responses.response import Response as OpenAIResponse
 from pydantic import BaseModel, ConfigDict, Field
-from pytest import fixture, param
+from pytest import fixture

 from agent_framework_azure_ai import AzureAIClient, AzureAISettings
 from agent_framework_azure_ai._shared import from_azure_ai_tools

-skip_if_azure_ai_integration_tests_disabled = pytest.mark.skipif(
-    os.getenv("AZURE_AI_PROJECT_ENDPOINT", "") in ("", "https://test-project.cognitiveservices.azure.com/")
-    or os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME", "") == "",
-    reason="No real AZURE_AI_PROJECT_ENDPOINT or AZURE_AI_MODEL_DEPLOYMENT_NAME provided; skipping integration tests.",
-)
-

@pytest.fixture
 def mock_project_client() -> MagicMock:
@@ -415,7 +407,7 @@ async def test_prepare_options_basic(mock_project_client: MagicMock) -> None:

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model"},
        ),
        patch.object(
@@ -452,7 +444,7 @@ async def test_prepare_options_with_application_endpoint(

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model"},
        ),
        patch.object(
@@ -494,7 +486,7 @@ async def test_prepare_options_with_application_project_client(

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model"},
        ),
        patch.object(
@@ -512,19 +504,6 @@ async def test_prepare_options_with_application_project_client(
        assert "extra_body" not in run_options


-async def test_initialize_client(mock_project_client: MagicMock) -> None:
-    """Test _initialize_client method."""
-    client = create_test_azure_ai_client(mock_project_client)
-
-    mock_openai_client = MagicMock()
-    mock_project_client.get_openai_client = MagicMock(return_value=mock_openai_client)
-
-    await client._initialize_client()
-
-    assert client.client is mock_openai_client
-    mock_project_client.get_openai_client.assert_called_once()
-
-
 def test_update_agent_name_and_description(mock_project_client: MagicMock) -> None:
    """Test _update_agent_name_and_description method."""
    client = create_test_azure_ai_client(mock_project_client)
@@ -827,14 +806,14 @@ async def test_runtime_tools_override_logs_warning(
    messages = [Message(role="user", contents=[Content.from_text(text="Hello")])]

    with patch(
-        "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+        "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
        return_value={"model": "test-model", "tools": [{"type": "function", "name": "tool_one"}]},
    ):
        await client._prepare_options(messages, {})

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model", "tools": [{"type": "function", "name": "tool_two"}]},
        ),
        patch("agent_framework_azure_ai._client.logger.warning") as mock_warning,
@@ -853,7 +832,7 @@ async def test_prepare_options_logs_warning_for_tools_with_existing_agent_versio

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model", "tools": [{"type": "function", "name": "tool_one"}]},
        ),
        patch("agent_framework_azure_ai._client.logger.warning") as mock_warning,
@@ -875,7 +854,7 @@ async def test_prepare_options_logs_warning_for_tools_on_application_endpoint(

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model", "tools": [{"type": "function", "name": "tool_one"}]},
        ),
        patch.object(client, "_get_agent_reference_or_create", new_callable=AsyncMock) as mock_get_agent_reference,
@@ -1101,14 +1080,14 @@ async def test_runtime_structured_output_override_logs_warning(
    messages = [Message(role="user", contents=[Content.from_text(text="Hello")])]

    with patch(
-        "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+        "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
        return_value={"model": "test-model"},
    ):
        await client._prepare_options(messages, {"response_format": ResponseFormatModel})

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={"model": "test-model"},
        ),
        patch("agent_framework_azure_ai._client.logger.warning") as mock_warning,
@@ -1129,7 +1108,7 @@ async def test_prepare_options_excludes_response_format(

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={
                "model": "test-model",
                "response_format": ResponseFormatModel,
@@ -1164,7 +1143,7 @@ async def test_prepare_options_keeps_values_for_unsupported_option_keys(

    with (
        patch(
-            "agent_framework.openai._responses_client.RawOpenAIResponsesClient._prepare_options",
+            "agent_framework_openai._chat_client.RawOpenAIChatClient._prepare_options",
            return_value={
                "model": "test-model",
                "tools": [{"type": "function", "name": "weather"}],
@@ -1365,352 +1344,6 @@ async def client() -> AsyncGenerator[AzureAIClient, None]:
            await project_client.agents.delete(agent_name=agent_name)


-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-@pytest.mark.parametrize(
-    "option_name,option_value,needs_validation",
-    [
-        # Simple ChatOptions - just verify they don't fail
-        param("top_p", 0.9, False, id="top_p"),
-        param("max_tokens", 500, False, id="max_tokens"),
-        param("seed", 123, False, id="seed"),
-        param("user", "test-user-id", False, id="user"),
-        param("metadata", {"test_key": "test_value"}, False, id="metadata"),
-        param("frequency_penalty", 0.5, False, id="frequency_penalty"),
-        param("presence_penalty", 0.3, False, id="presence_penalty"),
-        param("stop", ["END"], False, id="stop"),
-        param("allow_multiple_tool_calls", True, False, id="allow_multiple_tool_calls"),
-        param("tool_choice", "none", True, id="tool_choice_none"),
-        param("tool_choice", "auto", True, id="tool_choice_auto"),
-        param("tool_choice", "required", True, id="tool_choice_required_any"),
-        param(
-            "tool_choice",
-            {"mode": "required", "required_function_name": "get_weather"},
-            True,
-            id="tool_choice_required",
-        ),
-        # OpenAIResponsesOptions - just verify they don't fail
-        param("safety_identifier", "user-hash-abc123", False, id="safety_identifier"),
-        param("truncation", "auto", False, id="truncation"),
-        param("top_logprobs", 5, False, id="top_logprobs"),
-        param("prompt_cache_key", "test-cache-key", False, id="prompt_cache_key"),
-        param("max_tool_calls", 3, False, id="max_tool_calls"),
-    ],
-)
-async def test_integration_options(
-    option_name: str,
-    option_value: Any,
-    needs_validation: bool,
-    client: AzureAIClient,
-) -> None:
-    """Parametrized test covering options that can be set at runtime for a Foundry Agent.
-
-    Tests both streaming and non-streaming modes for each option to ensure
-    they don't cause failures. Options marked with needs_validation also
-    check that the feature actually works correctly.
-
-    This test reuses a single agent.
-    """
-    # Prepare test message
-    if option_name.startswith("tool_choice"):
-        # Use weather-related prompt for tool tests
-        messages = [Message(role="user", text="What is the weather in Seattle?")]
-    else:
-        # Generic prompt for simple options
-        messages = [Message(role="user", text="Say 'Hello World' briefly.")]
-
-    # Build options dict
-    options: dict[str, Any] = {option_name: option_value, "tools": [get_weather]}
-
-    for streaming in [False, True]:
-        if streaming:
-            # Test streaming mode
-            response_stream = client.get_response(
-                messages=messages,
-                stream=True,
-                options=options,
-            )
-
-            response = await response_stream.get_final_response()
-        else:
-            # Test non-streaming mode
-            response = await client.get_response(
-                messages=messages,
-                options=options,
-            )
-
-        assert response is not None
-        assert isinstance(response, ChatResponse)
-
-        # For tool_choice="required", we return after tool execution without a model text response
-        is_required_tool_choice = option_name == "tool_choice" and (
-            option_value == "required" or (isinstance(option_value, dict) and option_value.get("mode") == "required")
-        )
-
-        if is_required_tool_choice:
-            # Response should have function call and function result, but no text from model
-            assert len(response.messages) >= 2, f"Expected function call + result for {option_name}"
-            has_function_call = any(c.type == "function_call" for msg in response.messages for c in msg.contents)
-            has_function_result = any(c.type == "function_result" for msg in response.messages for c in msg.contents)
-            assert has_function_call, f"No function call in response for {option_name}"
-            assert has_function_result, f"No function result in response for {option_name}"
-        else:
-            assert response.text is not None, f"No text in response for option '{option_name}'"
-            assert len(response.text) > 0, f"Empty response for option '{option_name}'"
-
-        # Validate based on option type
-        if needs_validation:
-            if option_name.startswith("tool_choice") and not is_required_tool_choice:
-                # Should have called the weather function
-                text = response.text.lower()
-                assert "sunny" in text or "seattle" in text, f"Tool not invoked for {option_name}"
-            elif option_name == "response_format":
-                if option_value == OutputStruct:
-                    # Should have structured output
-                    assert response.value is not None, "No structured output"
-                    assert isinstance(response.value, OutputStruct)
-                    assert "seattle" in response.value.location.lower()
-                else:
-                    # Runtime JSON schema
-                    assert response.value is None, "No structured output, can't parse any json."
-                    response_value = json.loads(response.text)
-                    assert isinstance(response_value, dict)
-                    assert "location" in response_value
-                    assert "seattle" in response_value["location"].lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-@pytest.mark.parametrize(
-    "option_name,option_value,needs_validation",
-    [
-        param("temperature", 0.7, False, id="temperature"),
-        # Complex options requiring output validation
-        param("response_format", OutputStruct, True, id="response_format_pydantic"),
-        param(
-            "response_format",
-            {
-                "type": "json_schema",
-                "json_schema": {
-                    "name": "WeatherDigest",
-                    "strict": True,
-                    "schema": {
-                        "title": "WeatherDigest",
-                        "type": "object",
-                        "properties": {
-                            "location": {"type": "string"},
-                            "conditions": {"type": "string"},
-                            "temperature_c": {"type": "number"},
-                            "advisory": {"type": "string"},
-                        },
-                        "required": ["location", "conditions", "temperature_c", "advisory"],
-                        "additionalProperties": False,
-                    },
-                },
-            },
-            True,
-            id="response_format_runtime_json_schema",
-        ),
-    ],
-)
-async def test_integration_agent_options(
-    option_name: str,
-    option_value: Any,
-    needs_validation: bool,
-) -> None:
-    """Test Foundry agent level options in both streaming and non-streaming modes.
-
-    Tests both streaming and non-streaming modes for each option to ensure
-    they don't cause failures. Options marked with needs_validation also
-    check that the feature actually works correctly.
-
-    This test create a new client and uses it for both streaming and non-streaming tests.
-    """
-    async with temporary_chat_client(agent_name=f"test-agent-{option_name.replace('_', '-')}-{uuid4()}") as client:
-        for streaming in [False, True]:
-            # Prepare test message
-            if option_name.startswith("response_format"):
-                # Use prompt that works well with structured output
-                messages = [Message(role="user", text="The weather in Seattle is sunny")]
-                messages.append(Message(role="user", text="What is the weather in Seattle?"))
-            else:
-                # Generic prompt for simple options
-                messages = [Message(role="user", text="Say 'Hello World' briefly.")]
-
-            # Build options dict
-            options = {option_name: option_value}
-
-            if streaming:
-                # Test streaming mode
-                response_stream = client.get_response(
-                    messages=messages,
-                    stream=True,
-                    options=options,
-                )
-
-                response = await response_stream.get_final_response()
-            else:
-                # Test non-streaming mode
-                response = await client.get_response(
-                    messages=messages,
-                    options=options,
-                )
-
-            assert response is not None
-            assert isinstance(response, ChatResponse)
-            assert response.text is not None, f"No text in response for option '{option_name}'"
-            assert len(response.text) > 0, f"Empty response for option '{option_name}'"
-
-            # Validate based on option type
-            if needs_validation and option_name.startswith("response_format"):
-                if option_value == OutputStruct:
-                    # Should have structured output
-                    assert response.value is not None, "No structured output"
-                    assert isinstance(response.value, OutputStruct)
-                    assert "seattle" in response.value.location.lower()
-                else:
-                    # Runtime JSON schema
-                    assert response.value is None, "No structured output, can't parse any json."
-                    response_value = json.loads(response.text)
-                    assert isinstance(response_value, dict)
-                    assert "location" in response_value
-                    assert "seattle" in response_value["location"].lower()
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_web_search() -> None:
-    async with temporary_chat_client(agent_name="af-int-test-web-search") as client:
-        for streaming in [False, True]:
-            content = {
-                "messages": [
-                    Message(
-                        role="user",
-                        text="Who are the main characters of Kpop Demon Hunters? Do a web search to find the answer.",
-                    )
-                ],
-                "options": {
-                    "tool_choice": "auto",
-                    "tools": [client.get_web_search_tool()],
-                },
-            }
-            if streaming:
-                response = await client.get_response(stream=True, **content).get_final_response()
-            else:
-                response = await client.get_response(**content)
-
-            assert response is not None
-            assert isinstance(response, ChatResponse)
-            assert "Rumi" in response.text
-            assert "Mira" in response.text
-            assert "Zoey" in response.text
-
-            # Test that the client will use the web search tool with location
-            content = {
-                "messages": [
-                    Message(role="user", text="What is the current weather? Do not ask for my current location.")
-                ],
-                "options": {
-                    "tool_choice": "auto",
-                    "tools": [client.get_web_search_tool(user_location={"country": "US", "city": "Seattle"})],
-                },
-            }
-            if streaming:
-                response = await client.get_response(stream=True, **content).get_final_response()
-            else:
-                response = await client.get_response(**content)
-            assert response.text is not None
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_agent_hosted_mcp_tool() -> None:
-    """Integration test for MCP tool with Azure Response Agent using Microsoft Learn MCP."""
-    async with temporary_chat_client(agent_name="af-int-test-mcp") as client:
-        response = await client.get_response(
-            messages=[Message(role="user", text="How to create an Azure storage account using az cli?")],
-            options={
-                # this needs to be high enough to handle the full MCP tool response.
-                "max_tokens": 5000,
-                "tools": client.get_mcp_tool(
-                    name="Microsoft Learn MCP",
-                    url="https://learn.microsoft.com/api/mcp",
-                    description="A Microsoft Learn MCP server for documentation questions",
-                    approval_mode="never_require",
-                ),
-            },
-        )
-        assert isinstance(response, ChatResponse)
-        assert response.text
-        # Should contain Azure-related content since it's asking about Azure CLI
-        assert any(term in response.text.lower() for term in ["azure", "storage", "account", "cli"])
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_agent_hosted_code_interpreter_tool():
-    """Test Azure Responses Client agent with code interpreter tool through AzureAIClient."""
-    async with temporary_chat_client(agent_name="af-int-test-code-interpreter") as client:
-        response = await client.get_response(
-            messages=[Message(role="user", text="Calculate the sum of numbers from 1 to 10 using Python code.")],
-            options={
-                "tools": [client.get_code_interpreter_tool()],
-            },
-        )
-        # Should contain calculation result (sum of 1-10 = 55) or code execution content
-        contains_relevant_content = any(
-            term in response.text.lower() for term in ["55", "sum", "code", "python", "calculate", "10"]
-        )
-        assert contains_relevant_content or len(response.text.strip()) > 10
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_integration_agent_existing_session():
-    """Test Azure Responses Client agent with existing session to continue conversations across agent instances."""
-    # First conversation - capture the session
-    preserved_session = None
-
-    async with (
-        temporary_chat_client(agent_name="af-int-test-existing-session") as client,
-        Agent(
-            client=client,
-            instructions="You are a helpful assistant with good memory.",
-        ) as first_agent,
-    ):
-        # Start a conversation and capture the session
-        session = first_agent.create_session()
-        first_response = await first_agent.run("My hobby is photography. Remember this.", session=session, store=True)
-
-        assert isinstance(first_response, AgentResponse)
-        assert first_response.text is not None
-
-        # Preserve the session for reuse
-        preserved_session = session
-
-    # Second conversation - reuse the session in a new agent instance
-    if preserved_session:
-        async with (
-            temporary_chat_client(agent_name="af-int-test-existing-session-2") as client,
-            Agent(
-                client=client,
-                instructions="You are a helpful assistant with good memory.",
-            ) as second_agent,
-        ):
-            # Reuse the preserved session
-            second_response = await second_agent.run("What is my hobby?", session=preserved_session)
-
-            assert isinstance(second_response, AgentResponse)
-            assert second_response.text is not None
-            assert "photography" in second_response.text.lower()
-
-
 # region Factory Method Tests


@@ -2031,7 +1664,7 @@ async def test_inner_get_response_enriches_non_streaming(mock_project_client: Ma
    async def _fake_awaitable() -> ChatResponse:
        return base_response

-    with patch.object(RawOpenAIResponsesClient, "_inner_get_response", return_value=_fake_awaitable()):
+    with patch.object(RawOpenAIChatClient, "_inner_get_response", return_value=_fake_awaitable()):
        result_awaitable = client._inner_get_response(messages=[], options={}, stream=False)
        result = await result_awaitable  # type: ignore[misc]

@@ -2054,7 +1687,7 @@ async def test_inner_get_response_no_search_output_non_streaming(mock_project_cl
    async def _fake_awaitable() -> ChatResponse:
        return base_response

-    with patch.object(RawOpenAIResponsesClient, "_inner_get_response", return_value=_fake_awaitable()):
+    with patch.object(RawOpenAIChatClient, "_inner_get_response", return_value=_fake_awaitable()):
        result_awaitable = client._inner_get_response(messages=[], options={}, stream=False)
        result = await result_awaitable  # type: ignore[misc]

@@ -2075,7 +1708,7 @@ def test_inner_get_response_streaming_registers_hook(mock_project_client: MagicM

    mock_stream = _create_mock_stream()

-    with patch.object(RawOpenAIResponsesClient, "_inner_get_response", return_value=mock_stream):
+    with patch.object(RawOpenAIChatClient, "_inner_get_response", return_value=mock_stream):
        result = client._inner_get_response(messages=[], options={}, stream=True)

    assert result is mock_stream
@@ -2088,7 +1721,7 @@ def test_streaming_hook_captures_search_urls(mock_project_client: MagicMock) ->

    mock_stream = _create_mock_stream()

-    with patch.object(RawOpenAIResponsesClient, "_inner_get_response", return_value=mock_stream):
+    with patch.object(RawOpenAIChatClient, "_inner_get_response", return_value=mock_stream):
        client._inner_get_response(messages=[], options={}, stream=True)

    hook = mock_stream._transform_hooks[0]
@@ -2116,7 +1749,7 @@ def test_streaming_hook_enriches_url_citation(mock_project_client: MagicMock) ->

    mock_stream = _create_mock_stream()

-    with patch.object(RawOpenAIResponsesClient, "_inner_get_response", return_value=mock_stream):
+    with patch.object(RawOpenAIChatClient, "_inner_get_response", return_value=mock_stream):
        client._inner_get_response(messages=[], options={}, stream=True)

    hook = mock_stream._transform_hooks[0]
@@ -1,12 +1,10 @@
 # Copyright (c) Microsoft. All rights reserved.

-import os
 from unittest.mock import AsyncMock, MagicMock, patch

 import pytest
 from agent_framework import Agent, FunctionTool
 from agent_framework._mcp import MCPTool
-from azure.ai.projects.aio import AIProjectClient
 from azure.ai.projects.models import (
    AgentVersionDetails,
    PromptAgentDefinition,
@@ -14,16 +12,9 @@ from azure.ai.projects.models import (
 from azure.ai.projects.models import (
    FunctionTool as AzureFunctionTool,
 )
-from azure.identity.aio import AzureCliCredential

 from agent_framework_azure_ai import AzureAIProjectAgentProvider

-skip_if_azure_ai_integration_tests_disabled = pytest.mark.skipif(
-    os.getenv("AZURE_AI_PROJECT_ENDPOINT", "") in ("", "https://test-project.cognitiveservices.azure.com/")
-    or os.getenv("AZURE_AI_MODEL_DEPLOYMENT_NAME", "") == "",
-    reason="No real AZURE_AI_PROJECT_ENDPOINT or AZURE_AI_MODEL_DEPLOYMENT_NAME provided; skipping integration tests.",
-)
-

@pytest.fixture
 def mock_project_client() -> MagicMock:
@@ -689,42 +680,3 @@ async def test_provider_create_agent_with_mcp_and_regular_tools(
        assert "regular_function" in tool_names
        assert "mcp_function_1" in tool_names
        assert "mcp_function_2" in tool_names
-
-
-@pytest.mark.flaky
-@pytest.mark.integration
-@skip_if_azure_ai_integration_tests_disabled
-async def test_provider_create_and_get_agent_integration() -> None:
-    """Integration test for provider create_agent and get_agent."""
-    endpoint = os.environ["AZURE_AI_PROJECT_ENDPOINT"]
-    model = os.environ["AZURE_AI_MODEL_DEPLOYMENT_NAME"]
-
-    async with (
-        AzureCliCredential() as credential,
-        AIProjectClient(endpoint=endpoint, credential=credential) as project_client,
-    ):
-        provider = AzureAIProjectAgentProvider(project_client=project_client)
-
-        try:
-            # Create agent
-            agent = await provider.create_agent(
-                name="ProviderTestAgent",
-                model=model,
-                instructions="You are a helpful assistant. Always respond with 'Hello from provider!'",
-            )
-
-            assert isinstance(agent, Agent)
-            assert agent.name == "ProviderTestAgent"
-
-            # Run the agent
-            response = await agent.run("Hi!")
-            assert response.text is not None
-            assert len(response.text) > 0
-
-            # Get the same agent
-            retrieved_agent = await provider.get_agent(name="ProviderTestAgent")
-            assert retrieved_agent.name == "ProviderTestAgent"
-
-        finally:
-            # Cleanup
-            await project_client.agents.delete(agent_name="ProviderTestAgent")
@@ -13,10 +13,13 @@ from typing import Any, ClassVar, TypedDict
 from agent_framework import AGENT_FRAMEWORK_USER_AGENT, Message
 from agent_framework._sessions import BaseHistoryProvider
 from agent_framework._settings import SecretString, load_settings
-from agent_framework.azure._entra_id_authentication import AzureCredentialTypes
+from azure.core.credentials import TokenCredential
+from azure.core.credentials_async import AsyncTokenCredential
 from azure.cosmos import PartitionKey
 from azure.cosmos.aio import ContainerProxy, CosmosClient, DatabaseProxy

+AzureCredentialTypes = TokenCredential | AsyncTokenCredential
+
 logger = logging.getLogger(__name__)


@@ -14,7 +14,7 @@ cp .env.example .env

 Required variables:
 - `AZURE_OPENAI_ENDPOINT`
- `AZURE_OPENAI_CHAT_DEPLOYMENT_NAME`
+- `AZURE_OPENAI_DEPLOYMENT_NAME`
 - `AZURE_OPENAI_API_KEY`
 - `AzureWebJobsStorage`
 - `DURABLE_TASK_SCHEDULER_CONNECTION_STRING`
--- a/Show More
+++ b/Show More
Author	SHA1	Message	Date
Jacob AlberandGitHub	a35926518a	Merge branch 'main' into dev/dotnet_workflow/state_manager_object_check	2026-03-25 18:40:06 -04:00
Jacob Alber	e44be6611f	fix: Re-enable the "retrieve object" check in StateManager	2026-03-25 18:23:52 -04:00
Jacob AlberandGitHub	35f44e854e	.NET: fix: HandoffAgentExecutor does not output any response when non-streaming (#4745 ) * fix: HandoffAgentExecutor does not output any reponse when non-streaming * fix: Ensure Workflow outputs persisted in chat history when hosted AsAgent * fix: Remove duplicate history entry creation and ad test * test: Add streaming tests for AsAgent to smoke tests * feat: Add output configurability to Handoffs	2026-03-25 22:22:13 +00:00
Jacob AlberandGitHub	0756c45702	.NET: [BREAKING] Update type names and source generator to reduce conflicts (#4903 ) * refactor: [BREAKING] Config => ExecutorConfig Make the Config name less likely to collide with other classes by renaming to ExecutorConfig. Makes Configured and related classes internal as they do not need to be part of the public surface. * fix: Make RouteBuilder explicit in SourceGen to avoid conflicts	2026-03-25 20:35:17 +00:00
chetantoshniwal GitHub Copilot MAF Dashboard Bot westey	0b2ccd6126	.NET: Fix ChatOptions mutation in AIContextProviderChatClient across calls (#4891 ) * Fix UseAIContextProviders tool accumulation across calls (#4864) Clone ChatOptions before mutating it in InvokeProvidersAsync to prevent context provider tools from accumulating when the same ChatOptions instance is reused across multiple API calls. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Apply suggestions from code review Co-authored-by: chetantoshniwal <chetantoshniwal@gmail.com> * Apply suggestion from @westey-m --------- Co-authored-by: MAF Dashboard Bot <maf-dashboard-bot@users.noreply.github.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> Co-authored-by: westey <164392973+westey-m@users.noreply.github.com>	2026-03-25 20:02:16 +00:00
Peter IbekweandGitHub	23921c0f6e	.NET: Pass through external input request and handle response conversion for workflow as agent scenario (#4361 ) * Handle external input request and response conversion for workflow as agent scenario * Remove unnecessary test comment * Fix PR comments * Updated to fix edge cases, and add more tests. * Update pending requests to use typed properties instead of relying on StateBag. replying to PR feedback. * Fixed external response de-dup and updated possible brittle test. * Address PR comments on sending turn token for normal messages and handle contentId collision by source agent * Remove unnecessary serialization element and address pr comment on intercepted outgoing requests * Updated MEAI changes for UserInput request and response abstractions.	2026-03-25 19:23:44 +00:00
Peter Ibekwe GitHub Copilot	db16a9e74c	.NET: Clarify IResettableExecutor usage comment in workflow sample (#4905 ) * Clarify IResettableExecutor usage comment in workflow sample * Update dotnet/samples/03-workflows/Agents/WorkflowAsAnAgent/WorkflowFactory.cs Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Update dotnet/samples/03-workflows/Agents/WorkflowAsAnAgent/WorkflowFactory.cs Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>	2026-03-25 18:50:29 +00:00
Eduard van ValkenburgandGitHub	c012aac5f2	Python: [BREAKING] Reduce core dependencies and simplify optional integrations (#4904 ) * improved dependencies and some fixes * fix for mypy * improve mcp	2026-03-25 18:03:43 +00:00
Shyju KrishnankuttyandGitHub	49d69b3bf5	.NET: Expose workflows as MCP tools when hosting on Azure functions (#4768 ) * Expose workflow as MCP Tool * Expose workflow as MCP Tool * Cleanup * PR feedback fixes * update changelog to include PR numner * Improvements to error handling. * Adding a sample project demonstrating how to setup Agents and Workflows together. * Ensure duplicate agent registrations are properly handled.	2026-03-25 15:43:15 +00:00
Jacob AlberandGitHub	a9db40886e	fix: FS Checkpoint storage special character support (#4730 ) The `sessionId`, an optional parameter when starting a new session when running a workflow is an arbitrary string. This allows consumers to support whatever ids are needed by other systems, but can result in errors when an OS special or forbidden character is included. The fix is to escape the paths, in a 1:1 manner. We rely on EncodeDataString to do this. * Also modifies the index file to make it easier to determine what the name of the file on disk is for a given `sessionId`.	2026-03-25 15:21:30 +00:00
westeyandGitHub	87962e53c5	.NET: Persist messages during function call loop (#4762 ) * Persist messages during the Function Call Loop * Revert version reset * Fix bugs and improve sample * Fix formatting issues * Also updating conversation id during run * Update based on ADR feedback	2026-03-25 11:53:45 +00:00
Eduard van Valkenburg GitHub Copilot	5e056b672e	Python: [BREAKING] Python: Provider-leading client design & OpenAI package extraction (#4818 ) * Python: Provider-leading client design & OpenAI package extraction Major refactoring of the Python Agent Framework client architecture: - Extract OpenAI clients into new `agent-framework-openai` package - Core package no longer depends on openai, azure-identity, azure-ai-projects - Rename clients for discoverability: OpenAIResponsesClient → OpenAIChatClient, OpenAIChatClient → OpenAIChatCompletionClient - Unify `model_id`/`deployment_name`/`model_deployment_name` → `model` param - New FoundryChatClient for Azure AI Foundry Responses API - New FoundryAgent/FoundryAgentClient for connecting to pre-configured Foundry agents - Remove OpenAIBase/OpenAIConfigMixin from non-deprecated client MRO - Deprecate AzureOpenAI* clients, AzureAIClient, OpenAIAssistantsClient - Reorganize samples: azure_openai+azure_ai+azure_ai_agent → azure/ - ADR-0020: Provider-Leading Client Design Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: missing Agent imports in samples, .model_id → .model in foundry_local sample Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: CI failures — mypy errors, coverage targets, sample imports - azure-ai mypy: add type ignores for TypedDict total=, model arg, forward ref - Coverage: replace core.azure/openai targets with openai package target - project_provider: add type annotation for opts dict Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix: populate openai .pyi stub, fix broken README links, coverage targets Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fixes * updated observabilitty * reset azure init.pyi * fix errors * updated adr number * fix foundry local * fixed not renamed docstrings and comments, and added deprecated markers to old classes * fix tests and pyprojects * fix test vars * updated function tests * update durable * updated test setup for functions * Fix Foundry auth in workflow samples Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Stabilize Python integration workflows Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Update hosting samples for Foundry Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Trigger full CI rerun Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Trigger CI rerun again Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * trigger rerun * trigger rerun * fix for litellm * undo durabletask changes * Move Foundry APIs into foundry namespace Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Fix Foundry pyproject formatting Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Split provider samples by Foundry surface Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * Restore hosting sample requirements Also fix the Foundry Local sample link after the provider sample move. Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * updated tests * udpated foundry integration tests * removed dist from azurefunctions tests * Use separate Foundry clients for concurrent agents Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com> * fix client setup in azfunc and durable * disabled two tests * updated setup for some function and durable tests * improved azure openai setup with new clients * ignore deprecated * fixes * skip 11 * remove openai assistants int tests --------- Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>	2026-03-25 09:56:29 +00:00
Tao ChenandGitHub	4b533608b6	Python: Update sample validation scripts (#4870 ) * Update sample validation scripts * Adjust prompt * Update autogen-migration samples * Add fix suggestion * Split jobs * Add .env * Create trend report * Add timestamp * Add more env vars * Comments * force node24 * force node24 * force node22	2026-03-25 01:21:32 +00:00