What is the Model Context Protocol (MCP)?
The Model Context Protocol (MCP) is a shared plug that lets an AI assistant connect to the apps you already use. Anthropic released it in late 2024, and Claude, ChatGPT, Cursor, and others now speak it. Once an app like Gmail or Notion has an MCP connector, any MCP-aware assistant can read from it or take actions in it.
What MCP actually does
An AI assistant on its own only knows what it was trained on. It can't see your inbox, your spreadsheets, or your team's Notion pages. MCP is the standard that lets an assistant reach into those places when you ask it to. Anthropic published it as an open protocol in November 2024, and Claude Desktop, ChatGPT, Cursor, and VS Code now support it.
The official analogy is that MCP is a USB-C port for AI. Any AI tool that speaks MCP can plug into any app that ships an MCP connector, the same way any USB-C laptop works with any USB-C charger. Before this, every AI app built its own custom hookup to Gmail, its own to Slack, its own to your CRM. MCP replaces that with one shared shape.
What you can do with it
Real examples: an assistant pulling Google Calendar and Notion to plan your week, Claude Code generating a web app from a Figma file, an internal chatbot querying a CRM database without anyone exporting a spreadsheet first. Adding a new capability stops being a project: if a service has an MCP connector, turning it on takes minutes.
For developers
MCP is a client-server protocol over JSON-RPC. Servers expose three primitives: tools (actions), resources (readable data), and prompts (reusable templates). Servers run locally over stdio or remotely over HTTP with SSE. Spec and SDKs at modelcontextprotocol.io.
Last updated: May 20, 2026