Add turn item injection API (#17703)

## Summary
- Add `turn/inject_items` app-server v2 request support for appending
raw Responses API items to a loaded thread history without starting a
turn.
- Generate JSON schema and TypeScript protocol artifacts for the new
params and empty response.
- Document the new endpoint and include a request/response example.
- Preserve compatibility with the typo alias `turn/injet_items` while
returning the canonical method name.

## Testing
- Not run (not requested)

codex-rs/app-server/README.md

@@ -151,6 +151,7 @@ Example with notification opt-out:
 - `thread/backgroundTerminals/clean` — terminate all running background terminals for a thread (experimental; requires `capabilities.experimentalApi`); returns `{}` when the cleanup request is accepted.
 - `thread/rollback` — drop the last N turns from the agent’s in-memory context and persist a rollback marker in the rollout so future resumes see the pruned history; returns the updated `thread` (with `turns` populated) on success.
 - `turn/start` — add user input to a thread and begin Codex generation; responds with the initial `turn` object and streams `turn/started`, `item/*`, and `turn/completed` notifications. For `collaborationMode`, `settings.developer_instructions: null` means "use built-in instructions for the selected mode".
+- `thread/inject_items` — append raw Responses API items to a loaded thread’s model-visible history without starting a user turn; returns `{}` on success.
 - `turn/steer` — add user input to an already in-flight regular turn without starting a new turn; returns the active `turnId` that accepted the input. Review and manual compaction turns reject `turn/steer`.
 - `turn/interrupt` — request cancellation of an in-flight turn by `(thread_id, turn_id)`; success is an empty `{}` response and the turn finishes with `status: "interrupted"`.
 - `thread/realtime/start` — start a thread-scoped realtime session (experimental); returns `{}` and streams `thread/realtime/*` notifications. Omit `transport` for the websocket transport, or pass `{ "type": "webrtc", "sdp": "..." }` to create a WebRTC session from a browser-generated SDP offer; the remote answer SDP is emitted as `thread/realtime/sdp`.
@@ -581,6 +582,24 @@ Invoke a plugin by including a UI mention token such as `@sample` in the text in
 } } }
 ```
 
+### Example: Inject raw history items
+
+Use `thread/inject_items` to append prebuilt Responses API items to a loaded thread’s prompt history without starting a user turn. These items are persisted to the rollout and included in subsequent model requests.
+
+```json
+{ "method": "thread/inject_items", "id": 36, "params": {
+    "threadId": "thr_123",
+    "items": [
+        {
+            "type": "message",
+            "role": "assistant",
+            "content": [{ "type": "output_text", "text": "Previously computed context." }]
+        }
+    ]
+} }
+{ "id": 36, "result": {} }
+```
+
 ### Example: Start realtime with WebRTC
 
 Use `thread/realtime/start` with `transport.type: "webrtc"` when a browser or webview owns the `RTCPeerConnection` and app-server should create the server-side realtime session. The transport `sdp` must be the offer SDP produced by `RTCPeerConnection.createOffer()`, not a hand-written or minimal SDP string.

codex-rs/app-server/src/codex_message_processor.rs

@@ -132,6 +132,8 @@ use codex_app_server_protocol::ThreadForkParams;
 use codex_app_server_protocol::ThreadForkResponse;
 use codex_app_server_protocol::ThreadIncrementElicitationParams;
 use codex_app_server_protocol::ThreadIncrementElicitationResponse;
+use codex_app_server_protocol::ThreadInjectItemsParams;
+use codex_app_server_protocol::ThreadInjectItemsResponse;
 use codex_app_server_protocol::ThreadItem;
 use codex_app_server_protocol::ThreadListParams;
 use codex_app_server_protocol::ThreadListResponse;
@@ -958,6 +960,10 @@ impl CodexMessageProcessor {
                 )
                 .await;
             }
+            ClientRequest::ThreadInjectItems { request_id, params } => {
+                self.thread_inject_items(to_connection_request_id(request_id), params)
+                    .await;
+            }
             ClientRequest::TurnSteer { request_id, params } => {
                 self.turn_steer(to_connection_request_id(request_id), params)
                     .await;
@@ -6986,6 +6992,55 @@ impl CodexMessageProcessor {
         }
     }
 
+    async fn thread_inject_items(
+        &self,
+        request_id: ConnectionRequestId,
+        params: ThreadInjectItemsParams,
+    ) {
+        let (_, thread) = match self.load_thread(&params.thread_id).await {
+            Ok(value) => value,
+            Err(error) => {
+                self.outgoing.send_error(request_id, error).await;
+                return;
+            }
+        };
+
+        let items = match params
+            .items
+            .into_iter()
+            .enumerate()
+            .map(|(index, value)| {
+                serde_json::from_value::<ResponseItem>(value)
+                    .map_err(|err| format!("items[{index}] is not a valid response item: {err}"))
+            })
+            .collect::<std::result::Result<Vec<_>, _>>()
+        {
+            Ok(items) => items,
+            Err(message) => {
+                self.send_invalid_request_error(request_id, message).await;
+                return;
+            }
+        };
+
+        match thread.inject_response_items(items).await {
+            Ok(()) => {
+                self.outgoing
+                    .send_response(request_id, ThreadInjectItemsResponse {})
+                    .await;
+            }
+            Err(CodexErr::InvalidRequest(message)) => {
+                self.send_invalid_request_error(request_id, message).await;
+            }
+            Err(err) => {
+                self.send_internal_error(
+                    request_id,
+                    format!("failed to inject response items: {err}"),
+                )
+                .await;
+            }
+        }
+    }
+
     async fn set_app_server_client_info(
         thread: &CodexThread,
         app_server_client_name: Option<String>,

codex-rs/app-server/tests/common/mcp_process.rs

@@ -62,6 +62,7 @@ use codex_app_server_protocol::SkillsListParams;
 use codex_app_server_protocol::ThreadArchiveParams;
 use codex_app_server_protocol::ThreadCompactStartParams;
 use codex_app_server_protocol::ThreadForkParams;
+use codex_app_server_protocol::ThreadInjectItemsParams;
 use codex_app_server_protocol::ThreadListParams;
 use codex_app_server_protocol::ThreadLoadedListParams;
 use codex_app_server_protocol::ThreadMemoryModeSetParams;
@@ -602,6 +603,15 @@ impl McpProcess {
         self.send_request("turn/start", params).await
     }
 
+    /// Send a `thread/inject_items` JSON-RPC request (v2).
+    pub async fn send_thread_inject_items_request(
+        &mut self,
+        params: ThreadInjectItemsParams,
+    ) -> anyhow::Result<i64> {
+        let params = Some(serde_json::to_value(params)?);
+        self.send_request("thread/inject_items", params).await
+    }
+
     /// Send a `command/exec` JSON-RPC request (v2).
     pub async fn send_command_exec_request(
         &mut self,

codex-rs/app-server/tests/suite/v2/mod.rs

@@ -35,6 +35,7 @@ mod safety_check_downgrade;
 mod skills_list;
 mod thread_archive;
 mod thread_fork;
+mod thread_inject_items;
 mod thread_list;
 mod thread_loaded_list;
 mod thread_memory_mode_set;

codex-rs/app-server/tests/suite/v2/thread_inject_items.rs

@@ -0,0 +1,288 @@
+use anyhow::Context;
+use anyhow::Result;
+use app_test_support::McpProcess;
+use app_test_support::to_response;
+use codex_app_server_protocol::JSONRPCResponse;
+use codex_app_server_protocol::RequestId;
+use codex_app_server_protocol::ThreadInjectItemsParams;
+use codex_app_server_protocol::ThreadInjectItemsResponse;
+use codex_app_server_protocol::ThreadStartParams;
+use codex_app_server_protocol::ThreadStartResponse;
+use codex_app_server_protocol::TurnStartParams;
+use codex_app_server_protocol::UserInput as V2UserInput;
+use codex_core::RolloutRecorder;
+use codex_protocol::models::ContentItem;
+use codex_protocol::models::ResponseItem;
+use codex_protocol::protocol::InitialHistory;
+use codex_protocol::protocol::RolloutItem;
+use core_test_support::responses;
+use serde_json::Value;
+use std::path::Path;
+use tempfile::TempDir;
+use tokio::time::timeout;
+
+const DEFAULT_READ_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(10);
+
+#[tokio::test]
+async fn thread_inject_items_adds_raw_response_items_to_thread_history() -> Result<()> {
+    let server = responses::start_mock_server().await;
+    let body = responses::sse(vec![
+        responses::ev_response_created("resp-1"),
+        responses::ev_assistant_message("msg-1", "Done"),
+        responses::ev_completed("resp-1"),
+    ]);
+    let response_mock = responses::mount_sse_once(&server, body).await;
+
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let thread_req = mcp
+        .send_thread_start_request(ThreadStartParams {
+            model: Some("mock-model".to_string()),
+            ..Default::default()
+        })
+        .await?;
+    let thread_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
+    )
+    .await??;
+    let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
+
+    let injected_text = "Injected assistant context";
+    let injected_item = ResponseItem::Message {
+        id: None,
+        role: "assistant".to_string(),
+        content: vec![ContentItem::OutputText {
+            text: injected_text.to_string(),
+        }],
+        end_turn: None,
+        phase: None,
+    };
+
+    let inject_req = mcp
+        .send_thread_inject_items_request(ThreadInjectItemsParams {
+            thread_id: thread.id.clone(),
+            items: vec![serde_json::to_value(&injected_item)?],
+        })
+        .await?;
+    let inject_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(inject_req)),
+    )
+    .await??;
+    let _response: ThreadInjectItemsResponse =
+        to_response::<ThreadInjectItemsResponse>(inject_resp)?;
+
+    let rollout_path = thread.path.as_ref().context("thread path missing")?;
+    let history = RolloutRecorder::get_rollout_history(rollout_path).await?;
+    let InitialHistory::Resumed(resumed_history) = history else {
+        panic!("expected resumed rollout history");
+    };
+    assert!(
+        resumed_history
+            .history
+            .iter()
+            .any(|item| matches!(item, RolloutItem::ResponseItem(response_item) if response_item == &injected_item)),
+        "injected item should be persisted in rollout history"
+    );
+
+    let turn_req = mcp
+        .send_turn_start_request(TurnStartParams {
+            thread_id: thread.id.clone(),
+            input: vec![V2UserInput::Text {
+                text: "Hello".to_string(),
+                text_elements: Vec::new(),
+            }],
+            ..Default::default()
+        })
+        .await?;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(turn_req)),
+    )
+    .await??;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_notification_message("turn/completed"),
+    )
+    .await??;
+
+    let injected_value = serde_json::to_value(&injected_item)?;
+    let model_input = response_mock.single_request().input();
+    let environment_context_index =
+        response_item_text_position(&model_input, "<environment_context>")
+            .expect("environment context should be injected before the first user turn");
+    let injected_index = model_input
+        .iter()
+        .position(|item| item == &injected_value)
+        .expect("injected item should be sent in the next model request");
+    let user_prompt_index = response_item_text_position(&model_input, "Hello")
+        .expect("user prompt should be sent in the next model request");
+    assert!(
+        environment_context_index < injected_index,
+        "standard initial context should be sent before injected items"
+    );
+    assert!(
+        injected_index < user_prompt_index,
+        "injected items should be sent before the user prompt"
+    );
+
+    Ok(())
+}
+
+#[tokio::test]
+async fn thread_inject_items_adds_raw_response_items_after_a_turn() -> Result<()> {
+    let server = responses::start_mock_server().await;
+    let first_body = responses::sse(vec![
+        responses::ev_response_created("resp-1"),
+        responses::ev_assistant_message("msg-1", "First done"),
+        responses::ev_completed("resp-1"),
+    ]);
+    let second_body = responses::sse(vec![
+        responses::ev_response_created("resp-2"),
+        responses::ev_assistant_message("msg-2", "Second done"),
+        responses::ev_completed("resp-2"),
+    ]);
+    let response_mock = responses::mount_sse_sequence(&server, vec![first_body, second_body]).await;
+
+    let codex_home = TempDir::new()?;
+    create_config_toml(codex_home.path(), &server.uri())?;
+
+    let mut mcp = McpProcess::new(codex_home.path()).await?;
+    timeout(DEFAULT_READ_TIMEOUT, mcp.initialize()).await??;
+
+    let thread_req = mcp
+        .send_thread_start_request(ThreadStartParams {
+            model: Some("mock-model".to_string()),
+            ..Default::default()
+        })
+        .await?;
+    let thread_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(thread_req)),
+    )
+    .await??;
+    let ThreadStartResponse { thread, .. } = to_response::<ThreadStartResponse>(thread_resp)?;
+
+    let first_turn_req = mcp
+        .send_turn_start_request(TurnStartParams {
+            thread_id: thread.id.clone(),
+            input: vec![V2UserInput::Text {
+                text: "First turn".to_string(),
+                text_elements: Vec::new(),
+            }],
+            ..Default::default()
+        })
+        .await?;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(first_turn_req)),
+    )
+    .await??;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_notification_message("turn/completed"),
+    )
+    .await??;
+
+    let injected_item = ResponseItem::Message {
+        id: None,
+        role: "assistant".to_string(),
+        content: vec![ContentItem::OutputText {
+            text: "Injected after first turn".to_string(),
+        }],
+        end_turn: None,
+        phase: None,
+    };
+    let injected_value = serde_json::to_value(&injected_item)?;
+
+    let inject_req = mcp
+        .send_thread_inject_items_request(ThreadInjectItemsParams {
+            thread_id: thread.id.clone(),
+            items: vec![injected_value.clone()],
+        })
+        .await?;
+    let inject_resp: JSONRPCResponse = timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(inject_req)),
+    )
+    .await??;
+    let _response: ThreadInjectItemsResponse =
+        to_response::<ThreadInjectItemsResponse>(inject_resp)?;
+
+    let second_turn_req = mcp
+        .send_turn_start_request(TurnStartParams {
+            thread_id: thread.id.clone(),
+            input: vec![V2UserInput::Text {
+                text: "Second turn".to_string(),
+                text_elements: Vec::new(),
+            }],
+            ..Default::default()
+        })
+        .await?;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_response_message(RequestId::Integer(second_turn_req)),
+    )
+    .await??;
+    timeout(
+        DEFAULT_READ_TIMEOUT,
+        mcp.read_stream_until_notification_message("turn/completed"),
+    )
+    .await??;
+
+    let requests = response_mock.requests();
+    assert_eq!(requests.len(), 2);
+    assert!(
+        !requests[0].input().contains(&injected_value),
+        "injected item should not be sent before it is injected"
+    );
+    assert!(
+        requests[1].input().contains(&injected_value),
+        "injected item should be sent after being injected into existing history"
+    );
+
+    Ok(())
+}
+
+fn create_config_toml(codex_home: &Path, server_uri: &str) -> std::io::Result<()> {
+    let config_toml = codex_home.join("config.toml");
+    std::fs::write(
+        config_toml,
+        format!(
+            r#"
+model = "mock-model"
+approval_policy = "never"
+sandbox_mode = "read-only"
+
+model_provider = "mock_provider"
+
+[model_providers.mock_provider]
+name = "Mock provider for test"
+base_url = "{server_uri}/v1"
+wire_api = "responses"
+request_max_retries = 0
+stream_max_retries = 0
+"#
+        ),
+    )
+}
+
+fn response_item_text_position(items: &[Value], needle: &str) -> Option<usize> {
+    items.iter().position(|item| {
+        item.get("content")
+            .and_then(Value::as_array)
+            .into_iter()
+            .flatten()
+            .any(|content| {
+                content
+                    .get("text")
+                    .and_then(Value::as_str)
+                    .is_some_and(|text| text.contains(needle))
+            })
+    })
+}