MCP, the open plug

MCP is an open standard for connecting AI applications to tools and data. Think USB-C: one shape of plug, many devices.

Anthropic introduced the Model Context Protocol in November 2024. By 2026 it is the common language behind many connector catalogs. Claude, ChatGPT, Gemini, Grok, Cursor, and VS Code can all speak it. In December 2025 it moved under the Linux Foundation’s Agentic AI Foundation, which is a long way of saying: it is not one vendor’s private cable.

Before MCP, every assistant needed a custom integration for GitHub, then another for Postgres, then another for the internal wiki. After MCP, a team can expose one server and let many hosts use it.

Three roles

Host

The app you type into — Grok, Claude, ChatGPT, Cursor. It owns the conversation.

Client

A connector living inside the host. One client talks to one server. A host may run many clients at once.

Server

A program that offers tools (actions), resources (readable data), and sometimes prompt templates. It does not have to be the model.

The model never gets a raw password to your database. It sees a list of named tools with schemas — “search_issues,” “read_file,” “create_event.” It proposes a call. The host checks policy. The server executes. The result returns as text the model can read.

If you only ever press official catalog buttons, you can still understand MCP: those buttons are friendly faces on the same plug. If you build software, MCP is how you stop writing a one-off integration for every new chat product.