Practical MCP with FastMCP & LangChain

Why MCP Has Its Own Logging

If you have written Python before, you are probably familiar with the standard logging module. You create a logger, call logger.info("something happened"), and the message ends up in a file or on the console. That kind of logging is designed for the server operator — the person running the Python process. It writes to wherever the server's log output is configured to go. The MCP client never sees those messages.

MCP client logging is a different thing entirely. When your tool calls await ctx.info("Analyzing 42 dog breeds"), that message travels over the MCP protocol to the client application — Claude Desktop, a custom IDE integration, or whatever program connected to your server.

@mcp.tool
async def any_tool(ctx: Context) -> Dict[str, Any]:
    await