Model Context Protocol (MCP)
For engineers who want AI assistants and agents to use their company's data and tools through one open standard instead of a custom adapter per product. You will learn how the current, stateless MCP protocol works, build and secure servers over PostgreSQL, third-party APIs and event streams, and wire several of them into a tested research agent.
Course Content
What MCP is, how its messages and transports work under the stateless 2026-07-28 revision, how to build a server a model can actually use, and how to secure it. 3 lessons, about 50 minutes.
How to build MCP servers that sit safely in front of real systems: PostgreSQL and pgvector, rate-limited third-party APIs such as Google and Slack, and live event streams. 3 lessons, about 45 minutes.
Wire three MCP servers into one research assistant with a guarded agent loop, then build the test suite that keeps it reliable. 2 lessons, about 30 minutes.