What is MCP (Model Context Protocol)?

MCP, the Model Context Protocol, is an open standard for connecting AI apps to the outside world. Instead of every AI assistant needing a custom connector for every tool, MCP gives them one shared way to plug into data sources and services. Its official docs compare it to USB-C: one standard port that works everywhere.

How it works

MCP has three parts:

• Host: the AI app you use, such as Claude Desktop, ChatGPT or a code editor like VS Code.
• Client: the connection the host opens for each outside service.
• Server: a program that offers data or actions. It can run on your computer or in the cloud.

A server can offer tools (actions the AI can take, like querying a database), resources (data for context, like file contents) and prompts (reusable templates). Under the hood, messages use a standard format called JSON-RPC.

Who made it and who uses it

• November 2024: Anthropic open-sourced MCP with ready-made servers for Google Drive, Slack, GitHub and Postgres.
• December 2025: Anthropic donated it to the new Agentic AI Foundation under the Linux Foundation, co-founded with Block and OpenAI and backed by Google, Microsoft and AWS. At the time, Anthropic reported more than 10,000 active public MCP servers, and support in ChatGPT, Gemini, Microsoft Copilot, Cursor and VS Code.

Why it matters

MCP is a big reason AI agents can now do real work across apps: a developer builds a connection once, and it works with many AI assistants.

The risks

Connecting AI to real tools means real access. The MCP spec warns that it opens paths to data access and code execution, and says apps must ask for your consent before sharing data or running tools. The protocol can't enforce that by itself, so safety depends on each app, and on connecting only servers you trust.