Welcome to Model Context Protocol (MCP), the fourth module of the Generative AI & LLM Engineering course. In the last module you learned to give a model tools by hand — defining a schema, running the loop, returning results. That works, but every app reinvents it, and a tool you build for one project can’t be reused in another. MCP fixes that. It is an open standard — think “USB-C for AI tools” — that lets any AI application talk to any tool or data source through one common protocol.

You’ll learn what MCP actually is and why a standard matters, then connect to an MCP server from Python and discover its tools. You’ll build your own server with FastMCP — turning plain Python functions into MCP tools — and finally bridge MCP to Claude, so the model can use tools served by MCP exactly the way it used your hand-written tools in Module 3. The capstone wires Claude to a small service you build.

Every example runs for real: you’ll start a server, connect a client, and watch Claude call an MCP-provided tool and answer with the result. Because MCP separates who provides a tool from who uses it, the server you write here could later be used by Claude Desktop, an IDE, or any other MCP-compatible app — without changing a line.

Start with Lesson 1, where you’ll learn what MCP is and the problem it was built to solve.