Using FastAPI with Streamable HTTP MCP servers

[vietnv1602]

when I use Streamable HTTP MCP servers inside a FastAPI router, I encounter an issue with an asyncio task.

server = MCPServerStreamableHttp()
await server.connect()
...
await server.cleanup()  --> this line always causes an asyncio error

How can i fix this

[seratch]

This looks like a general FastAPI question, but you may want to use the lifecycle events to cleanup the MCP servers you start when bootstrapping the whole app. https://fastapi.tiangolo.com/advanced/events/#shutdown-event

[vietnv1602]

@seratch here is my code

@router.post("/chat")
async def chat(...):
  mcp_word = MCPServerStreamableHttp(
          name="Word",
          params={
              "url": settings.MCP_WORD,
              "headers": {"X-Conversation-Id": conversation_id},
          },
          cache_tools_list=True,
          max_retry_attempts=3,
          client_session_timeout_seconds=300,
      )
  await mcp_word.connect()
  agent = Agent(
                name="Agent Word",
                model=agent_model,
                mcp_servers=[mcp_word],
                instructions=(
                    "You are a specialized agent for handling Microsoft Word (.docx) files."
                    "You can read, edit, create, summarize, extract and convert Word documents while preserving formatting."
                    "Always act precisely according to user instructions, never delete or add content without explicit request."
                ),
            )
  ...
  response = Runner.run_streamed(
        starting_agent=agent,
        input=messages,
        run_config=RunConfig(
            tracing_disabled=True, session_input_callback=session_input_callback
        ),
        session=session,
    )
    return StreamingResponse(
        stream_response(response),
        media_type="text/event-stream",
    )

when the response ends, I get an error:

Task exception was never retrieved
future: <Task finished name='Task-79' coro=<<async_generator_athrow without __name__>()> exception=RuntimeError('Attempted to exit cancel scope in a different task than it was entered in')>
  + Exception Group Traceback (most recent call last):
  |   File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 781, in __aexit__
  |     raise BaseExceptionGroup(
  |         "unhandled errors in a TaskGroup", self._exceptions
  |     ) from None
  | BaseExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\mcp\client\streamable_http.py", line 502, in streamablehttp_client
    |     yield (
    |     ...<3 lines>...
    |     )
    | GeneratorExit
    +------------------------------------

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\mcp\client\streamable_http.py", line 478, in streamablehttp_client
    async with anyio.create_task_group() as tg:
               ~~~~~~~~~~~~~~~~~~~~~~~^^
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 787, in __aexit__
    if self.cancel_scope.__exit__(type(exc), exc, exc.__traceback__):
       ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 459, in __exit__
    raise RuntimeError(
    ...<2 lines>...
    )
RuntimeError: Attempted to exit cancel scope in a different task than it was entered in
Task exception was never retrieved
future: <Task finished name='Task-78' coro=<<async_generator_athrow without __name__>()> exception=RuntimeError('Attempted to exit cancel scope in a different task than it was entered in')>
  + Exception Group Traceback (most recent call last):
  |   File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 781, in __aexit__
  |     raise BaseExceptionGroup(
  |         "unhandled errors in a TaskGroup", self._exceptions
  |     ) from None
  | BaseExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\mcp\client\streamable_http.py", line 502, in streamablehttp_client
    |     yield (
    |     ...<3 lines>...
    |     )
    | GeneratorExit
    +------------------------------------

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\mcp\client\streamable_http.py", line 478, in streamablehttp_client
    async with anyio.create_task_group() as tg:
               ~~~~~~~~~~~~~~~~~~~~~~~^^
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 787, in __aexit__
    if self.cancel_scope.__exit__(type(exc), exc, exc.__traceback__):
       ~~~~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "D:\gitlab\ai-server-xbot\venv\Lib\site-packages\anyio\_backends\_asyncio.py", line 459, in __exit__
    raise RuntimeError(
    ...<2 lines>...
    )
RuntimeError: Attempted to exit cancel scope in a different task than it was entered in

[seratch]

if you need to spin an MCP server up for each request, your stream handing code is responsible to shut down the server when the streaming is done.

[vietnv1602]

is the issue in the Agent SDK?
When I use Jupyter Notebook with the code:

from agents.mcp import MCPServerStreamableHttp

mcp_word = MCPServerStreamableHttp(
        name="excel server",
        params={
            "url": "http://localhost:8000/mcp",
            "timeout": 10,
            "headers": {"X-Conversation-Id": "test-async"},
        },
        cache_tools_list=True,
        max_retry_attempts=3,
        client_session_timeout_seconds=300,
    )
await mcp_word.connect()


await mcp_word.cleanup()

i got an error: Error cleaning up server: Attempted to exit cancel scope in a different task than it was entered in

[seratch]

I don't think so. I've verified it works without issues by modifying one of the examples to run in a way similar to your code:

diff --git a/examples/mcp/prompt_server/main.py b/examples/mcp/prompt_server/main.py
index 4caa95d8..b7a53103 100644
--- a/examples/mcp/prompt_server/main.py
+++ b/examples/mcp/prompt_server/main.py
@@ -73,16 +73,18 @@ async def show_available_prompts(mcp_server: MCPServer):
 
 
 async def main():
-    async with MCPServerStreamableHttp(
+    server = MCPServerStreamableHttp(
         name="Simple Prompt Server",
         params={"url": "http://localhost:8000/mcp"},
-    ) as server:
-        trace_id = gen_trace_id()
-        with trace(workflow_name="Simple Prompt Demo", trace_id=trace_id):
-            print(f"Trace: https://platform.openai.com/traces/trace?trace_id={trace_id}\n")
-
-            await show_available_prompts(server)
-            await demo_code_review(server)
+    )
+    await server.connect()
+    trace_id = gen_trace_id()
+    with trace(workflow_name="Simple Prompt Demo", trace_id=trace_id):
+        print(f"Trace: https://platform.openai.com/traces/trace?trace_id={trace_id}\n")
+
+        await show_available_prompts(server)
+        await demo_code_review(server)
+    await server.cleanup()
 
 
 if __name__ == "__main__":

I haven't checked what's wrong with your FastAPI app settings, but there may be something specific to your app project.

[github-actions]

This issue is stale because it has been open for 7 days with no activity.