/agents/mcp/server.py", line 115, in list_tools raise UserError("Server not initialized. Make sure you call `connect()` first.") agents.exceptions.UserError: Server not initialized. Make sure you call `connect()` first.

[fengyang95]

Please read this first

• Have you read the docs?Agents SDK docs
• Have you searched for related issues? Others may have had similar requests

Question

Using the following code, why does it throw an error? Is there a convenient way to integrate MCP with agent orchestration?

  SERVER_FILE_SYSTEM = MCPServerStdio(
    name="Filesystem Server, via npx",
    params={
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-filesystem", WORK_SPACE],
    },
)

SEARCHER_AGENT = Agent(
    name="Searcher",
    instructions="Search for related files and content",
    handoff_description="""Responsible for finding relevant files and code content based on given information""",
    model=ModelConfig.cr_ultra_model,
    mcp_servers=[
        SERVER_FILE_SYSTEM
    ]
)

if __name__ == "__main__":
    from agents import Runner

    async def main():
        with trace("File search example"):
            result = await Runner.run(starting_agent=SEARCHER_AGENT, input="List directory structure")
            print(result)

    asyncio.run(main())

[DanieleMorotti]

Hi, even if you didn't specify the error, it is likely due to how you pass the server to the agent. As shown in the example, using the context manager automatically calls connect().

For instance, try the following code:

import asyncio

from agents import (
    Agent,
    Runner
)
from agents.mcp import MCPServerStdio


SERVER_FILE_SYSTEM = MCPServerStdio(
    name="Filesystem Server, via npx",
    params={
        "command": "npx",
        "args": ["-y", "@modelcontextprotocol/server-filesystem", WORK_SPACE],
    },
)

if __name__ == "__main__":
    from agents import Runner

    async def main():
        async with SERVER_FILE_SYSTEM as server:
            SEARCHER_AGENT = Agent(
                name="Searcher",
                instructions="Search for related files and content",
                handoff_description="""Responsible for finding relevant files and code content based on given information""",
                model=ModelConfig.cr_ultra_model,
                mcp_servers=[server]
            )
            result = await Runner.run(starting_agent=SEARCHER_AGENT, input="List directory structure")
            print(result)

    asyncio.run(main())

This worked for me.

[rm-openai]

That's right ^

Either use with server, or manually call server.connect() and server.cleanup()