_handle_stateless_request ClosedResourceError

[281743556]

Initial Checks

• I confirm that I'm using the latest version of MCP Python SDK
• I confirm that I searched for my issue in https://github.com/modelcontextprotocol/python-sdk/issues before opening this issue

Description

Every request will try an exception

use _handle_stateless_request > run_stateless_server > http_transport.connect() > for session_message in write_stream_reader
It will go in twice, and an exception(yio.ClosedResourceError) will be thrown when the second request has been closed

Example Code

app = Server("mcp-server")
def main(port: int, transport: str, dev: bool, limit_concurrency: int) -> int:
    # Create the session manager with true stateless mode
    from starlette.applications import Starlette
    from starlette.routing import Mount, Route
    session_manager = StreamableHTTPSessionManager(
        app=app,
        event_store=None,
        json_response=True,
        stateless=True, 
    )

    @contextlib.asynccontextmanager
    async def lifespan(app: Starlette) -> AsyncIterator[None]:
        """Context manager for managing session manager lifecycle."""
        async with session_manager.run():
            logger.info("Application started with StreamableHTTP session manager!")
            try:
                yield
            finally:
                logger.info("Application shutting down...")

    async def handle_streamable_http(
        scope: Scope, receive: Receive, send: Send
    ) -> None:
        await session_manager.handle_request(scope, receive, send)

    starlette_app = Starlette(
        debug=dev,
        routes=[
            Mount("/message", app=handle_streamable_http), 
        ],
        lifespan=lifespan,
    )
    import uvicorn
    uvicorn.run(starlette_app, host="0.0.0.0", port=port, http="httptools", limit_concurrency=limit_concurrency)

if __name__ == "__main__":
    asyncio.run(main())

view logs:

[2025-07-31 15:10:41,884][MainThread:19320][task_id:mcp.server.streamable_http][streamable_http.py:630][INFO][Terminating session: None]
[2025-07-31 15:10:41,885][MainThread:19320][task_id:mcp.server.streamable_http][streamable_http.py:880][ERROR][Error in message router]
Traceback (most recent call last):
  File "d:\project\bpaas\apihub-mcp-server\.venv\Lib\site-packages\mcp\server\streamable_http.py", line 831, in message_router
    async for session_message in write_stream_reader:
    ...<46 lines>...
            )
  File "d:\project\bpaas\apihub-mcp-server\.venv\Lib\site-packages\anyio\abc\_streams.py", line 35, in __anext__
    return await self.receive()
           ^^^^^^^^^^^^^^^^^^^^
  File "d:\project\bpaas\apihub-mcp-server\.venv\Lib\site-packages\anyio\streams\memory.py", line 111, in receive
    return self.receive_nowait()
           ~~~~~~~~~~~~~~~~~~~^^
  File "d:\project\bpaas\apihub-mcp-server\.venv\Lib\site-packages\anyio\streams\memory.py", line 93, in receive_nowait
    raise ClosedResourceError
anyio.ClosedResourceError

Terminating has already released memory first

debug for : first

debug for : second

And the log shows that Terminating should be executed first, followed by create_task_group

Python & MCP Python SDK

MCP version: 1.12.2
Python：3.13

[WYBangin]

我也遇到了一模一样的问题, 目前只有使用1.11.0版本的mcp才能解决问题，是否有其他解决办法:
MCP version: 1.12.0
anyio version: 4.7.0
Python: 3.13.5

[Spikhalskiy]

One more version of similar exception from MPC Server with streamable-http transport:

  + Exception Group Traceback (most recent call last):
  |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/server/streamable_http_manager.py", line 241, in run_server
  |     await self.app.run(
  |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/server/lowlevel/server.py", line 614, in run
  |     async with AsyncExitStack() as stack:
  |   File "/opt/conda/envs/my-env/lib/python3.11/contextlib.py", line 745, in __aexit__
  |     raise exc_details[1]
  |   File "/opt/conda/envs/my-env/lib/python3.11/contextlib.py", line 728, in __aexit__
  |     cb_suppress = await cb(*exc_details)
  |                   ^^^^^^^^^^^^^^^^^^^^^^
  |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/shared/session.py", line 218, in __aexit__
  |     return await self._task_group.__aexit__(exc_type, exc_val, exc_tb)
  |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/anyio/_backends/_asyncio.py", line 781, in __aexit__
  |     raise BaseExceptionGroup(
  | ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Exception Group Traceback (most recent call last):
    |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/server/lowlevel/server.py", line 625, in run
    |     async with anyio.create_task_group() as tg:
    |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/anyio/_backends/_asyncio.py", line 781, in __aexit__
    |     raise BaseExceptionGroup(
    | ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
    +-+---------------- 1 ----------------
      | Traceback (most recent call last):
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/server/lowlevel/server.py", line 648, in _handle_message
      |     await self._handle_request(message, req, session, lifespan_context, raise_exceptions)
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/server/lowlevel/server.py", line 712, in _handle_request
      |     await message.respond(response)
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/shared/session.py", line 131, in respond
      |     await self._session._send_response(  # type: ignore[reportPrivateUsage]
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/mcp/shared/session.py", line 329, in _send_response
      |     await self._write_stream.send(session_message)
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/anyio/streams/memory.py", line 243, in send
      |     self.send_nowait(item)
      |   File "/opt/conda/envs/my-env/lib/python3.11/site-packages/anyio/streams/memory.py", line 212, in send_nowait
      |     raise ClosedResourceError
      | anyio.ClosedResourceError
      +------------------------------------

Observation: MPC client receives the response (that an error occured), so it's not connected a network issue.
Python 3.11.13
MCP 1.18.0
MCP Client from llama-index-tools-mcp 0.4.2

[Spikhalskiy]

Just wanted to share for people who have MCP Servers in production and are looking for a solution for this long-standing problem, https://github.com/jlowin/fastmcp doesn't have the same issue. Migration is painless, at least for simple setups - renaming imports from mpc to fastmpc will likely be enough.

[ofek]

Reproducible example here #1190 (comment)