Model Context Protocol

The Model Context Protocol (MCP) is an open standard that defines how an AI application talks to outside tools and data. Anthropic published it in November 2024, and it was soon adopted across the industry, including by OpenAI, Google and Microsoft. Before MCP, every AI app needed its own custom integration for every service it wanted to use; with MCP, a tool built once works with any app that speaks the protocol.

It has two sides. An MCP server is a small program that wraps some system (your files, a calendar, a company database, GitHub, a design tool) and describes what it offers: tools the AI can call to do things, resources it can read, and ready-made prompts. An MCP client lives inside the AI app, such as Claude, ChatGPT or a code editor, connects to servers, and shows the model what is available. When you ask 'what is on my calendar tomorrow?', the model decides to call the calendar server's tool, gets the answer, and uses it in its reply. Servers can run on your own computer or remotely over the internet.

That makes MCP a building block for AI agents, which need to act, not just talk. It also deserves care. A server can do whatever its permissions allow, so a careless or malicious one could read private data or take actions you did not intend, and text coming back from a tool can carry hidden instructions meant to trick the model (prompt injection). Install servers only from sources you trust, give them the minimum access they need, and keep an eye on what an assistant asks to do.