This sample demonstrates how to use a local MCP (Model Context Protocol) client with Azure Foundry Agents. Unlike the hosted MCP approach where Azure Foundry invokes the MCP server on the service side, this sample connects to the MCP server directly from the client via HTTP (Streamable HTTP transport) and passes the resolved tools to the agent.

What this sample demonstrates

Connecting to an MCP server locally using HttpClientTransport
Discovering available tools from the MCP server client-side
Passing locally-resolved MCP tools to a Foundry agent
Using the Microsoft Learn MCP endpoint for documentation search
Managing agent lifecycle (creation and deletion)

Prerequisites

Before you begin, ensure you have the following prerequisites:

.NET 10 SDK or later
Azure Foundry service endpoint and deployment configured
Azure CLI installed and authenticated (for Azure credential authentication)

Note: This demo uses Azure CLI credentials for authentication. Make sure you're logged in with az login and have access to the Azure Foundry resource. For more information, see the Azure CLI documentation.

Set the following environment variables:

$env:AZURE_FOUNDRY_PROJECT_ENDPOINT="https://your-foundry-service.services.ai.azure.com/api/projects/your-foundry-project" # Replace with your Azure Foundry resource endpoint
$env:AZURE_FOUNDRY_PROJECT_DEPLOYMENT_NAME="gpt-4o-mini"  # Optional, defaults to gpt-4o-mini

Run the sample

Navigate to the FoundryAgents sample directory and run:

cd dotnet/samples/GettingStarted/FoundryAgents
dotnet run --project .\FoundryAgents_Step27_LocalMCP

Expected behavior

The sample will:

Connect to the Microsoft Learn MCP server via HTTP and list available tools
Create an agent with the locally-resolved MCP tools
Ask two questions about Microsoft documentation
The agent will use the MCP tools (invoked locally) to search Microsoft Learn documentation
Display the agent's responses with information from the documentation
Clean up resources by deleting the agent