What is MCP server?

An MCP server is a program that exposes tools, resources, and prompts to AI agents over the Model Context Protocol, giving a model a uniform way to read data or take actions in an external system.

An MCP server is the provider side of the Model Context Protocol. It wraps some capability, a database, a search engine, a code host, a memory store, and exposes it through the protocol's vocabulary of tools (actions the model can call), resources (data it can read), and prompts (reusable templates). The server declares what it offers during initialization, then responds to the client's calls. Servers run in one of two main modes: locally as a subprocess that communicates over stdio, or remotely as an HTTP endpoint the client connects to over the network, often behind OAuth. A well-designed server keeps its tool surface small and well-described so the model can choose the right tool, and it scopes credentials tightly because anything it can do, a prompted agent can do. Examples in this directory range from the official GitHub server to community database and chat servers.