MCPServerStdio bug when passing env vars as param

[fcestari]

Initial Checks

• I confirm that I'm using the latest version of Pydantic AI

Description

Hi. I am having issues to start an mcp server with the class MCPServerStdio whenever I pass a dict[str, str] to the env parameter.

I can confirm that I have npx installed and have used with success the MCPServerStdio without passing the env parameter.

Example Code

import asyncio

from pydantic_ai import Agent
from pydantic_ai.models.openai import OpenAIModel
from pydantic_ai.providers.openai import OpenAIProvider
from pydantic_ai.mcp import MCPServerStdio

github_server = MCPServerStdio(
    command="npx",
    args=["@modelcontextprotocol/server-github"],
    env={"GITHUB_PERSONAL_ACCESS_TOKEN": "<github-pat-here>"},
)

ollama_model = OpenAIModel(
    model_name="llama3.2", provider=OpenAIProvider(base_url="http://localhost:11434/v1")
)

agent = Agent(ollama_model, mcp_servers=[github_server])


async def main():
    async with agent.run_mcp_servers():
        result = await agent.run("list issues from repo XXX")
    print(result.data)


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

---
Error:

File "/opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/subprocess.py", line 1974, in _execute_child
    raise child_exception_type(errno_num, err_msg, err_filename)
FileNotFoundError: [Errno 2] No such file or directory: 'npx'

Python, Pydantic AI & LLM client version

3.13, 0.0.43, openai v1.68.0 (azure openai and ollama)

[Kludex]

It was fixed on mcp's side: modelcontextprotocol/python-sdk#327

If you bump the mcp dependency to 1.5.0, it should work as expected (the release is in progress, should be available in some hours).

[fcestari]

Thanks!