Add close_sse_stream callback to tool context (per findleyr feedback)

Addresses feedback from TypeScript SDK PR #1129 on the SEP-1699 implementation:

1. Reduces coupling between tools and transport by providing a callback
   instead of requiring direct session_manager access
2. Makes event store awareness transparent - callback is None if no
   event store is configured (events would be lost otherwise)
3. Adds per-call retry_interval parameter for tool-specific reconnection
   timing

Changes:
- Add CloseSSEStreamCallback type to mcp/shared/context.py
- Add close_sse_stream field to RequestContext (lowlevel API)
- Add close_sse_stream property to FastMCP Context (high-level API)
- Add retry_interval parameter to close_sse_stream methods
- Create callback factory in StreamableHTTPServerTransport
- Inject callback via ServerMessageMetadata
- Update example servers to use new callback API

Author: felixweinberger

examples/servers/everything-server/mcp_everything_server/server.py

@@ -36,7 +36,6 @@
     TextResourceContents,
 )
 from pydantic import AnyUrl, BaseModel, Field
-from starlette.requests import Request
 
 logger = logging.getLogger(__name__)
 
@@ -315,13 +314,11 @@ async def test_reconnection(ctx: Context[ServerSession, None]) -> str:
     # Send notification before disconnect
     await ctx.info("Notification before disconnect")
 
-    # Get session_id from request headers
-    request = ctx.request_context.request
-    if isinstance(request, Request):
-        session_id = request.headers.get("mcp-session-id")
-        if session_id:
-            # Trigger server-initiated SSE disconnect
-            await mcp.session_manager.close_sse_stream(session_id, ctx.request_id)
+    # Use the close_sse_stream callback if available
+    # This is None if not on streamable HTTP transport or no event store configured
+    if ctx.close_sse_stream:
+        # Trigger server-initiated SSE disconnect with optional retry interval
+        await ctx.close_sse_stream(retry_interval=3000)  # 3 seconds
 
     # Wait for client to reconnect
     await asyncio.sleep(0.2)

examples/snippets/servers/sse_polling_server.py

@@ -24,7 +24,6 @@
 
 import anyio
 from starlette.applications import Starlette
-from starlette.requests import Request
 from starlette.routing import Mount
 from starlette.types import Receive, Scope, Send
 
@@ -89,9 +88,6 @@ def create_app() -> Starlette:
     """Create the Starlette application with SSE polling example server."""
     app = Server("sse-polling-example")
 
-    # Store reference to session manager for close_sse_stream access
-    session_manager_ref: list[StreamableHTTPSessionManager] = []
-
     @app.call_tool()
     async def call_tool(name: str, arguments: dict[str, Any]) -> list[types.ContentBlock]:
         if name != "long-task":
@@ -121,15 +117,11 @@ async def call_tool(name: str, arguments: dict[str, Any]) -> list[types.ContentB
         await anyio.sleep(1)
 
         # Server-initiated disconnect - client will reconnect
-        if session_manager_ref:
+        # Use the close_sse_stream callback if available
+        # This is None if not on streamable HTTP transport or no event store configured
+        if ctx.close_sse_stream:
             logger.info(f"[{request_id}] Closing SSE stream to trigger polling reconnect...")
-            session_manager = session_manager_ref[0]
-            # Get session ID from the request and close the stream via public API
-            request = ctx.request
-            if isinstance(request, Request):
-                session_id = request.headers.get("mcp-session-id")
-                if session_id:
-                    await session_manager.close_sse_stream(session_id, request_id)
+            await ctx.close_sse_stream(retry_interval=2000)  # 2 seconds
 
         # Wait a bit for client to reconnect
         await anyio.sleep(0.5)
@@ -175,7 +167,6 @@ async def list_tools() -> list[types.Tool]:
         # Tell clients to reconnect after 2 seconds
         retry_interval=2000,
     )
-    session_manager_ref.append(session_manager)
 
     async def handle_mcp(scope: Scope, receive: Receive, send: Send) -> None:
         await session_manager.handle_request(scope, receive, send)

src/mcp/server/fastmcp/server.py

@@ -64,7 +64,7 @@
 from mcp.server.streamable_http import EventStore
 from mcp.server.streamable_http_manager import StreamableHTTPSessionManager
 from mcp.server.transport_security import TransportSecuritySettings
-from mcp.shared.context import LifespanContextT, RequestContext, RequestT
+from mcp.shared.context import CloseSSEStreamCallback, LifespanContextT, RequestContext, RequestT
 from mcp.types import Annotations, AnyFunction, ContentBlock, GetPromptResult, Icon, ToolAnnotations
 from mcp.types import Prompt as MCPPrompt
 from mcp.types import PromptArgument as MCPPromptArgument
@@ -1287,6 +1287,23 @@ def session(self):
         """Access to the underlying session for advanced usage."""
         return self.request_context.session
 
+    @property
+    def close_sse_stream(self) -> CloseSSEStreamCallback | None:
+        """Callback to close SSE stream for polling behavior (SEP-1699).
+
+        This allows tools to trigger server-initiated SSE disconnect during
+        long-running operations, enabling client reconnection with polling.
+
+        Returns None if:
+        - Not running on streamable HTTP transport
+        - No event store configured (events would be lost)
+
+        Usage:
+            if ctx.close_sse_stream:
+                await ctx.close_sse_stream(retry_interval=3000)  # Reconnect after 3s
+        """
+        return self.request_context.close_sse_stream
+
     # Convenience methods for common log levels
     async def debug(self, message: str, **extra: Any) -> None:
         """Send a debug log message."""

src/mcp/server/lowlevel/server.py

@@ -680,12 +680,14 @@ async def _handle_request(
 
             token = None
             try:
-                # Extract request context from message metadata
+                # Extract request context and callback from message metadata
                 request_data = None
+                close_sse_stream_callback = None
                 if message.message_metadata is not None and isinstance(
                     message.message_metadata, ServerMessageMetadata
                 ):  # pragma: no cover
                     request_data = message.message_metadata.request_context
+                    close_sse_stream_callback = message.message_metadata.close_sse_stream
 
                 # Set our global state that can be retrieved via
                 # app.get_request_context()
@@ -696,6 +698,7 @@ async def _handle_request(
                         session,
                         lifespan_context,
                         request=request_data,
+                        close_sse_stream=close_sse_stream_callback,
                     )
                 )
                 response = await handler(req)

src/mcp/server/streamable_http.py

@@ -28,6 +28,7 @@
     TransportSecurityMiddleware,
     TransportSecuritySettings,
 )
+from mcp.shared.context import CloseSSEStreamCallback
 from mcp.shared.message import ServerMessageMetadata, SessionMessage
 from mcp.shared.version import SUPPORTED_PROTOCOL_VERSIONS
 from mcp.types import (
@@ -283,6 +284,26 @@ async def _create_priming_event(self, stream_id: str) -> dict[str, str | int] |
 
         return event_data
 
+    def _create_close_sse_stream_callback(self, request_id: RequestId) -> CloseSSEStreamCallback | None:
+        """Create a bound callback for closing SSE streams.
+
+        Args:
+            request_id: The request ID to bind to the callback
+
+        Returns:
+            A callback that closes the SSE stream for this request,
+            or None if no event store is configured (events would be lost).
+        """
+        # Only provide callback if event store is configured
+        # Without an event store, closing the stream would lose events
+        if self._event_store is None:
+            return None
+
+        async def callback(retry_interval: int | None = None) -> bool:
+            return await self.close_sse_stream(request_id, retry_interval)
+
+        return callback
+
     async def _clean_up_memory_streams(self, request_id: RequestId) -> None:  # pragma: no cover
         """Clean up memory streams for a given request ID."""
         if request_id in self._request_streams:
@@ -544,7 +565,12 @@ async def sse_writer():
                     async with anyio.create_task_group() as tg:
                         tg.start_soon(response, scope, receive, send)
                         # Then send the message to be processed by the server
-                        metadata = ServerMessageMetadata(request_context=request)
+                        # Create callback for closing SSE stream (only if event store configured)
+                        close_callback = self._create_close_sse_stream_callback(request_id)
+                        metadata = ServerMessageMetadata(
+                            request_context=request,
+                            close_sse_stream=close_callback,
+                        )
                         session_message = SessionMessage(message, metadata=metadata)
                         await writer.send(session_message)
                 except Exception:
@@ -716,26 +742,35 @@ async def terminate(self) -> None:
             # During cleanup, we catch all exceptions since streams might be in various states
             logger.debug(f"Error closing streams: {e}")
 
-    async def close_sse_stream(self, request_id: RequestId) -> None:
+    async def close_sse_stream(self, request_id: RequestId, retry_interval: int | None = None) -> bool:
         """Close an SSE stream for a specific request, triggering client reconnection.
 
         Use this to implement polling behavior during long-running operations -
         client will reconnect after the retry interval specified in the priming event.
 
         Args:
             request_id: The request ID (or stream key) of the stream to close
+            retry_interval: Optional retry interval in ms to send before closing.
+                           If provided, overrides the transport's default retry interval.
+
+        Returns:
+            True if the stream was found and closed, False otherwise.
         """
         request_id_str = str(request_id)
-        if request_id_str in self._request_streams:
-            try:
-                sender, receiver = self._request_streams[request_id_str]
-                await sender.aclose()
-                await receiver.aclose()
-            except Exception:  # pragma: no cover
-                # Stream might already be closed
-                logger.debug(f"Error closing SSE stream {request_id_str} - may already be closed")
-            finally:
-                self._request_streams.pop(request_id_str, None)
+        if request_id_str not in self._request_streams:
+            return False
+
+        try:
+            sender, receiver = self._request_streams[request_id_str]
+            await sender.aclose()
+            await receiver.aclose()
+            return True
+        except Exception:  # pragma: no cover
+            # Stream might already be closed
+            logger.debug(f"Error closing SSE stream {request_id_str} - may already be closed")
+            return False
+        finally:
+            self._request_streams.pop(request_id_str, None)
 
     async def _handle_unsupported_request(self, request: Request, send: Send) -> None:  # pragma: no cover
         """Handle unsupported HTTP methods."""

src/mcp/server/streamable_http_manager.py

@@ -282,7 +282,7 @@ async def run_server(*, task_status: TaskStatus[None] = anyio.TASK_STATUS_IGNORE
             await response(scope, receive, send)
 
     async def close_sse_stream(  # pragma: no cover
-        self, session_id: str, request_id: str | int
+        self, session_id: str, request_id: str | int, retry_interval: int | None = None
     ) -> bool:
         """Close an SSE stream for a specific request, triggering client reconnection.
 
@@ -292,12 +292,13 @@ async def close_sse_stream(  # pragma: no cover
         Args:
             session_id: The MCP session ID (from mcp-session-id header)
             request_id: The request ID of the stream to close
+            retry_interval: Optional retry interval in ms to send before closing.
+                           If provided, overrides the transport's default retry interval.
 
         Returns:
             True if the stream was found and closed, False otherwise
         """
         if session_id not in self._server_instances:
             return False
         transport = self._server_instances[session_id]
-        await transport.close_sse_stream(request_id)
-        return True
+        return await transport.close_sse_stream(request_id, retry_interval)

src/mcp/shared/context.py

@@ -1,5 +1,5 @@
 from dataclasses import dataclass
-from typing import Any, Generic
+from typing import Any, Generic, Protocol
 
 from typing_extensions import TypeVar
 
@@ -11,10 +11,27 @@
 RequestT = TypeVar("RequestT", default=Any)
 
 
+class CloseSSEStreamCallback(Protocol):
+    """Callback to close SSE stream for polling behavior (SEP-1699).
+
+    Args:
+        retry_interval: Optional retry interval in ms to send before closing.
+                       If None, uses the transport's default retry interval.
+
+    Returns:
+        True if the stream was found and closed, False otherwise.
+    """
+
+    async def __call__(self, retry_interval: int | None = None) -> bool: ...
+
+
 @dataclass
 class RequestContext(Generic[SessionT, LifespanContextT, RequestT]):
     request_id: RequestId
     meta: RequestParams.Meta | None
     session: SessionT
     lifespan_context: LifespanContextT
     request: RequestT | None = None
+    # Callback to close SSE stream for polling behavior (SEP-1699)
+    # None if not on streamable HTTP transport or no event store configured
+    close_sse_stream: CloseSSEStreamCallback | None = None

src/mcp/shared/message.py

@@ -7,9 +7,13 @@
 
 from collections.abc import Awaitable, Callable
 from dataclasses import dataclass
+from typing import TYPE_CHECKING
 
 from mcp.types import JSONRPCMessage, RequestId
 
+if TYPE_CHECKING:
+    from mcp.shared.context import CloseSSEStreamCallback
+
 ResumptionToken = str
 
 ResumptionTokenUpdateCallback = Callable[[ResumptionToken], Awaitable[None]]
@@ -30,6 +34,9 @@ class ServerMessageMetadata:
     related_request_id: RequestId | None = None
     # Request-specific context (e.g., headers, auth info)
     request_context: object | None = None
+    # Callback to close SSE stream for polling behavior (SEP-1699)
+    # None if not on streamable HTTP transport or no event store configured
+    close_sse_stream: "CloseSSEStreamCallback | None" = None
 
 
 MessageMetadata = ClientMessageMetadata | ServerMessageMetadata | None