MCP Sample (HTTP SSE mode) Failed with HTTP 503 Error

[neolee]

Initial Checks

• I confirm that I'm using the latest version of Pydantic AI
• I confirm that I searched for my issue in https://github.com/pydantic/pydantic-ai/issues before opening this issue

Description

Trying official example here: https://ai.pydantic.dev/mcp/client/

The stdio mode sample runs without any issue but the sse mode sample failed with error:

  + Exception Group Traceback (most recent call last):
  |   File "/Users/neo/Code/ML/neb/mcp_client.py", line 26, in <module>
  |     asyncio.run(run(agent_sse))
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 190, in run
  |     return runner.run(main)
  |            ^^^^^^^^^^^^^^^^
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/runners.py", line 118, in run
  |     return self._loop.run_until_complete(task)
  |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete
  |     return future.result()
  |            ^^^^^^^^^^^^^^^
  |   File "/Users/neo/Code/ML/neb/mcp_client.py", line 18, in run
  |     async with agent.run_mcp_servers():
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/contextlib.py", line 210, in __aenter__
  |     return await anext(self.gen)
  |            ^^^^^^^^^^^^^^^^^^^^^
  |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/pydantic_ai/agent.py", line 1299, in run_mcp_servers
  |     await exit_stack.enter_async_context(mcp_server)
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/contextlib.py", line 650, in enter_async_context
  |     result = await _enter(cm)
  |              ^^^^^^^^^^^^^^^^
  |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/pydantic_ai/mcp.py", line 86, in __aenter__
  |     self._read_stream, self._write_stream = await self._exit_stack.enter_async_context(self.client_streams())
  |                                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/contextlib.py", line 650, in enter_async_context
  |     result = await _enter(cm)
  |              ^^^^^^^^^^^^^^^^
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/contextlib.py", line 210, in __aenter__
  |     return await anext(self.gen)
  |            ^^^^^^^^^^^^^^^^^^^^^
  |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/pydantic_ai/mcp.py", line 219, in client_streams
  |     async with sse_client(
  |   File "/opt/homebrew/Cellar/python@3.11/3.11.11/Frameworks/Python.framework/Versions/3.11/lib/python3.11/contextlib.py", line 210, in __aenter__
  |     return await anext(self.gen)
  |            ^^^^^^^^^^^^^^^^^^^^^
  |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/mcp/client/sse.py", line 43, in sse_client
  |     async with anyio.create_task_group() as tg:
  |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/anyio/_backends/_asyncio.py", line 772, in __aexit__
  |     raise BaseExceptionGroup(
  | ExceptionGroup: unhandled errors in a TaskGroup (1 sub-exception)
  +-+---------------- 1 ----------------
    | Traceback (most recent call last):
    |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/mcp/client/sse.py", line 53, in sse_client
    |     event_source.response.raise_for_status()
    |   File "/Users/neo/Code/ML/neb/.venv/lib/python3.11/site-packages/httpx/_models.py", line 829, in raise_for_status
    |     raise HTTPStatusError(message, request=request, response=self)
    | httpx.HTTPStatusError: Server error '503 Service Unavailable' for url 'http://localhost:3001/sse'

The server is already up with command:

❯ npx @pydantic/mcp-run-python sse
Running MCP server with SSE transport on localhost:3001

Any hint?

Example Code

from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerHTTP

server_sse = MCPServerHTTP(url='http://localhost:3001/sse')
agent_sse = Agent('openai:gpt-4o', mcp_servers=[server_sse])

async def run(agent):
    async with agent.run_mcp_servers():
        result = await agent.run('How many days between 2000-01-01 and 2025-03-28?')
    print(result.data)


if __name__ == "__main__":
    import asyncio

    asyncio.run(run(agent_sse))

Python, Pydantic AI & LLM client version

Python: 3.11.11
pydantic: 2.11.0
pydantic-ai: 0.0.46
openai: 1.69.0

[neolee]

Additional info: curl test seems OK:

❯ curl -v http://localhost:3001/sse
* Host localhost:3001 was resolved.
* IPv6: ::1
* IPv4: 127.0.0.1
*   Trying [::1]:3001...
* Connected to localhost (::1) port 3001
> GET /sse HTTP/1.1
> Host: localhost:3001
> User-Agent: curl/8.7.1
> Accept: */*
>
* Request completely sent off
< HTTP/1.1 200 OK
< X-Powered-By: Express
< Content-Type: text/event-stream
< Cache-Control: no-cache
< Connection: keep-alive
< Date: Wed, 02 Apr 2025 12:30:29 GMT
< Transfer-Encoding: chunked
<
event: endpoint
data: /messages?sessionId=6dc2f13b-d3dc-4834-a920-539abdcb3f23

[samuelcolvin]

I don't really understand what's going on here. as you say the server is find with curl if you get this 503 again, happy to try to fix, but without more information, I'm not sure what I can do.

[neolee]

@samuelcolvin what kind of info do you need?

To clarify, I've run the server by npx @pydantic/mcp-run-python sse and the curl -v http://localhost:3001/sse test seems OK (as shown above), but as the same time the client code sample (copy your official sample from here) return 503 error (seems not connected to the server). This is beyond my understanding 😂

[neolee]

@samuelcolvin tried to dig deeper and found something. The mcp-run-python server only listens at the IPv6 address ::1 and the client created using MCPServerHTTP(url='http://localhost:3001/sse') seems only connect to IPv4 address (I tried another MCP server which listens IPv4 address and it works). I don't know why but that is the symptoms on my system.

p.s. I'm working on a MBP with macOS Sequoia 15.3.2 and M1 chip.

[samuelcolvin]

I have the same machine and it works fine.

As it happens, we're switching to deno from npx anyway, please try #1340 and see if that fixes your problem.

[neolee]

@samuelcolvin switching to deno does fix my issue. The @pydantic/mcp-run-python server running by deno does accept IPv4 connection and client code runs pretty well. Still don't get it but at least it runs. Thanks for your advice any way.

[github-actions]

This issue is stale, and will be closed in 3 days if no reply is received.

[billaram]

Have you tried running it under async with agent.run_mcp_servers(): context managers?