تشغيل الخادم باستخدام stdio

You have a server object and a tool. Now you need to actually run it so a client can connect and use what you built. 🏃

The simplest transport is stdio: standard input and output. The client launches your script and talks to it through those two streams.

Starting the server is one method. Calling mcp.run() hands control to FastMCP, which begins listening for messages right away.

Pass a transport argument to be explicit. For local use you want "stdio", which is also the default when you call run.

Wrap the call in the classic __main__ guard. That way the server starts only when the file is run directly, not when imported.

Here is a complete server: import, instance, tool, and run. Just a handful of lines and it speaks MCP.

One golden rule: never print to standard output. That stream carries protocol messages, and stray text will corrupt them.

Need to debug? Write to stderr, not stdout. The client ignores stderr, so your notes stay out of the protocol stream.

With stdio, you rarely run the script by hand. The host app starts your process for you using a configured command.

After run() the server blocks, waiting for messages. It isn't frozen; it's listening for the next request from the client.

Remember that stdio works only on the same machine, since it relies on a launched process. Remote servers use HTTP transport instead, which comes later.

Why must you avoid plain print() in a stdio server?

You ran the server with mcp.run() over stdio, guarded it with __main__, and kept stdout clean. It's live and listening. Let's confirm it works! ✅