Tools are how an agent acts on the world: read a database, call an API, open a ticket. The Model Context Protocol (MCP) is an open standard for offering those tools to any model or client. An MCP server describes its tools, resources and prompts; a client such as an agent discovers them at runtime and calls them over a transport like stdio or streamable HTTP.
This cluster covers both sides. It starts with what a server exposes and how an agent connects to one, then builds a production MCP server in Python, turns an existing FastAPI app or OpenAPI spec into a server, and adds authentication with JWT and OAuth. The later guides deal with running servers for real: security practices, testing with pytest, keeping the tool list small enough for the model to choose well, caching and circuit breakers, and when a plain CLI is the better choice.
The code in these guides runs on Promptise Foundry, an open-source Python framework for AI agents and MCP servers.