An MCP server is a small program that offers tools, data or prompt templates to AI apps through the Model Context Protocol, a shared standard for connecting them.
In plain words
MCP is an open standard, often compared to a USB-C port, that lets any compatible AI app plug into outside systems the same way. A server can offer three kinds of things: tools the model may call with the user's approval, resources such as file contents to read, and ready-made prompts. You write the functions, the SDK describes them for the model, and an AI app connects to your server.
A quick example
The official tutorial builds a weather server with two tools, one for weather alerts in a US state and one for a forecast, then connects it to Claude for Desktop. A tip from the guide: a server that talks over standard input and output must not print to stdout, so log to stderr.
Where you'll meet it
- Connecting assistants to company data and APIs.
- Giving coding agents access to tools such as browsers.
Related
Sources
- DocsBuild an MCP server, Model Context Protocol docs · read 29 Sept 2026
- DocsWhat is the Model Context Protocol (MCP)?, Model Context Protocol docs · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 29 Sept 2026. Spotted an error? Report it.