You have likely witnessed the hype-cycle around MCP (the Model-Context Protocol) for LLMs. It was heralded as \"the universal interface between LLMs and the world\" but then faded into the background as attention shifted towards AI Agents. Yet, the background of your AI app is exactly where an MCP should be, and in this talk we cover why. We will tour the MCP protocol, the Python reference implementation, and an example agent using an MCP. Expect protocol flow-charts, architecture diagrams, and a real-world demo. You will walk away knowing the core ideas of MCP, how it connects to the broader ecosystem, and how to power your AI agents.