The Model Context Protocol (MCP) is the shared standard that lets AI assistants use outside systems: your data, your APIs, your expertise. You build a server once, and any MCP client can use it. In this track you'll build a weather server with FastMCP 4, the Python framework that turns ordinary functions into MCP tools with one decorator. You'll add all three MCP primitives (tools, resources, and prompts), test the server in memory, run it over HTTP, and deploy it to a public URL on Prefect Horizon where you can chat with it. Everything runs in a browser sandbox. You only need a GitHub account for the final deployment challenge.
Write a one-tool MCP server in a dozen lines of Python and call it from the command line.
Decorate three working weather functions and read the contract the assistant receives.
Give the assistant something to read and a saved recipe to run, using the other two MCP primitives.
Use FastMCP's in-memory client to test the server the way an assistant would, in under a second.
Run your server over HTTP, push it to GitHub, deploy it to a public URL, and chat with it.
