logfire.instrument_mcp

[samuelcolvin]

Description

Idea is to instrument:

• client messages
• SSE messages from the server
• record logs as individual zero-duration spans? at least on the client

Needs:

• Client hangs when ctx.info() logging with pure Python client modelcontextprotocol/python-sdk#201
• add span around tool calls pydantic-ai#1158 (for best use with pydantic-ai)

Also needs POST messages to be shown with the right trace/parent span.

E.g. if I run the following I see:

from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerHTTP

import logfire

logfire.configure()
logfire.instrument_httpx()
logfire.instrument_pydantic_ai()

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


async def main():
    async with agent.run_mcp_servers():
        result = await agent.run('How many days between 2000-01-01 and 2025-03-18?')
    print(result.data)
    #> There are 9,208 days between January 1, 2000, and March 18, 2025.


if __name__ == '__main__':
    import asyncio

    asyncio.run(main())

I see:

All those posts to localhost/messages should be within the agent run trace

[Kludex]

I think using instrument_starlette should already give something for SSE.