Build an MCP server with Python

Prerequisites

Step-by-Step

1. 1

   Install FastMCP

   FastMCP is included in the official mcp Python SDK. Use uv for the snappiest install.

   uv add 'mcp[cli]'

2. 2

   Create the server

   Decorate functions with @mcp.tool. The framework introspects type hints and docstrings to build the schema.

   from mcp.server.fastmcp import FastMCP
   
   mcp = FastMCP('weather')
   
   @mcp.tool()
   def get_weather(city: str) -> str:
       """Return current weather for the given city."""
       return f'It is 72F in {city}'

3. 3

   Add a resource

   Resources are read-only context the agent can pull. Great for config, recent logs, or canned data.

   @mcp.resource('config://app')
   def config() -> str:
       return open('config.yaml').read()

4. 4

   Run it

   The CLI provides stdio and HTTP transports. Use stdio for local clients.

   uv run mcp run server.py

5. 5

   Inspect with the dev tool

   mcp dev launches the inspector against your server.

   uv run mcp dev server.py

6. 6

   Connect from a client

   Add the server to your client config and reload.

   {
     "mcpServers": {
       "weather": { "command": "uv", "args": ["run", "mcp", "run", "/abs/path/server.py"] }
     }
   }

Common Pitfalls

What's Next

• ->Add @mcp.prompt for reusable templates.
• ->Deploy via SSE or HTTP transport for remote use.

More Build an MCP server