René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"The cleanup procedure after \"yield\" in lifespan is unreachable on Windows","articleBody":"### Initial Checks\n\n- [x] I confirm that I'm using the latest version of MCP Python SDK\n- [x] I confirm that I searched for my issue in https://github.com/modelcontextprotocol/python-sdk/issues before opening this issue\n\n### Description\n\nDear developpers,\n\nI am developing with MCP SDK Python on Windows. I noticed that the resource cleanup process of my local process defined inside the `lifespan` is not executed at all. To investigate the issue, I tested with the minimal code below and found that the code after the `yield` statement is never executed, regardless of whether an error occurs or not.\nI haven’t been able to test this on platforms other than Windows or with the SSE transport.\n\n```python\n\"\"\"agent_runner.py\n\nA minimum code to use MCP server.\nPlease run this file.\n\n\"\"\"\nimport os\nimport sys\nimport asyncio\nimport logging\nfrom dotenv import load_dotenv\nfrom openai import AsyncOpenAI\nfrom agents import Agent, Runner, OpenAIChatCompletionsModel\nfrom agents.mcp import MCPServerStdio, MCPServerStdioParams\n\nload_dotenv()  # loading OPENAI_API_KEY\n\nlogging.basicConfig(level=logging.INFO)\nlogger = logging.getLogger(\"agent_runner\")\n\n# disable library loggers\nlogging.getLogger('mcp').setLevel(logging.ERROR)\nlogging.getLogger('httpx').setLevel(logging.ERROR)\n\n\nasync def main():\n\n    # Launch MCP server\n    params = MCPServerStdioParams(\n        command=sys.executable,\n        args=[os.path.join(os.path.dirname(__file__), \"srv_stdio.py\")],\n        env={},\n        encoding=\"utf-8\"\n    )\n    async with MCPServerStdio(params=params, name=\"MCP tool\") as mcp_server:\n\n        logger.info(\"4. MCP stdio server running\")\n\n        # Create agent\n        agent = Agent(\n            name=\"AgentWithMCP\",\n            instructions=\"Answer with using tools.\",\n            model=OpenAIChatCompletionsModel(model=\"gpt-4\", openai_client=AsyncOpenAI()),\n            mcp_servers=[mcp_server]\n        )\n\n        prompt = \"Please reverse the word 'hello'.\"\n        logger.info(f\"  PROMPT: {prompt}\")\n        result = await Runner.run(agent, input=prompt)\n        logger.info(f\"  FINAL OUTPUT: {result.final_output}\")\n\n\nif __name__ == \"__main__\":\n    logger.info(\"1. Starting program...\")\n    asyncio.run(main())\n    logger.info(\"8. Program terminated.\")\n```\n\n```python\n\n\"\"\"srv_stdio.py\n\nA minimum code to show ignoring after \"yield\" in lifetime.\nPlease put this file on the same folder as \"agent_runner.py\".\n\n\"\"\"\nfrom __future__ import annotations as _annotations\n\nimport os\nimport sys\nimport logging\nfrom collections.abc import AsyncIterator\nfrom contextlib import asynccontextmanager\nfrom mcp.server.fastmcp import FastMCP\nfrom mcp.server.lowlevel.server import Server, LifespanResultT, RequestT\n\nlogging.basicConfig(level=logging.INFO)\nlogger = logging.getLogger(\"stdio server\")\n\n# disable library loggers\nlogging.getLogger('mcp').setLevel(logging.ERROR)\nlogging.getLogger('httpx').setLevel(logging.ERROR)\n\nif sys.platform == \"win32\" and os.environ.get('PYTHONIOENCODING') is None:\n    sys.stdin.reconfigure(encoding=\"utf-8\")\n    sys.stdout.reconfigure(encoding=\"utf-8\")\n    sys.stderr.reconfigure(encoding=\"utf-8\")\n\n\n@asynccontextmanager\nasync def lifespan(server: Server[LifespanResultT, RequestT]) -\u003e AsyncIterator[object]:\n    logger.info('3. Launching Server...')\n    try:\n        yield {}\n    finally:\n        logger.info('6. Terminating Server '\n                    '(Want to release my resources here).')  # not shown\n\n\nmcp = FastMCP(\n    \"EchoMCPServer\",\n    lifespan=lifespan,\n)\n\n\n@mcp.tool()\ndef echo(msg: str) -\u003e str:\n    \"\"\"Make the passed string reversed.\"\"\"\n    logger.info(f\"5. [echo] called with msg='{msg}'\")\n    return msg[::-1]\n\n\nif __name__ == \"__main__\":\n    logger.info(\"2. Starting MCP server...\")\n    mcp.run()\n    logger.info(\"7. MCP server terminated.\")  # not shown\n```\n\nThe outputs missing \"6. Terminating Server \\~\" and \"7. MCP server terminated.\".\n```\nINFO:agent_runner:1. Starting program...\nINFO:stdio server:2. Starting MCP server...\nINFO:stdio server:3. Launching Server...\nINFO:agent_runner:4. MCP stdio server running\nINFO:agent_runner:  PROMPT: Please reverse the word 'hello'.\nINFO:stdio server:5. [echo] called with msg='olleh'\nINFO:agent_runner:  FINAL OUTPUT: The reverse of 'hello' is 'olleh'.\nINFO:agent_runner:8. Program terminated.\n```\n\nI suspected that this issue might be due to the process being forcibly terminated when exiting the `with` block in `MCPServerStdio`. Upon investigation, I found that modifying the following section allows the sample code to correctly display \"6. Terminating Server \\~\" and \"7. MCP server terminated.\", and the cleanup process is executed as intended.\n\n```python\n# mcp/client/stdio/win32.py\n102  -         process.terminate()\n102  +         os.kill(process.pid, signal.CTRL_C_EVENT)\n```\n\nThis change seems to resolve the issue on my end, but do you think it could be helpful for improving the MCP SDK overall?\n\n### Example Code\n\n```Python\n\n```\n\n### Python \u0026 MCP Python SDK\n\n```Text\nPython 3.10\nmcp 1.9.4\nopenai-agents 0.0.17\n```","author":{"url":"https://github.com/ScatterTemple","@type":"Person","name":"ScatterTemple"},"datePublished":"2025-06-25T16:30:54.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":3},"url":"https://github.com/1027/python-sdk/issues/1027"}