AI

MCP

MCP (Model Context Protocol): an open protocol that connects AI agents to external tools and data sources in a standardized way.

Cinq fiches de connexion en verre transparent aux formes différentes convergeant en éventail vers un unique port bordeaux lumineux, sur un socle circulaire en métal noir, sous un faisceau blanc sur fond sombre
MCP — Image générée par IA (Qwen-Image 2.1, ComfyUI local)

MCP (Model Context Protocol) is an open protocol that lets AI agents connect to tools and data sources in a standardized way. Created by Anthropic in November 2024, it became the de facto standard for agent interoperability within a year — adopted by OpenAI, Google, Microsoft and thousands of developers.

Technical definition

The problem MCP solves

Without MCP, every agent has to be wired by hand to every tool it uses. An agent that needs to read a Figma file, query a database and send an email requires three specific integrations, each with its own format, authentication and call logic. Multiply that by the number of tools and agents in an organization and you get a combinatorial problem.

MCP solves it with a client-server approach:

  • The MCP client (the AI agent) speaks the MCP protocol
  • The MCP server (the tool) exposes its capabilities in a standardized format
  • The agent automatically discovers what the server can do and uses it

The USB analogy

MCP is to AI agents what USB is to computer peripherals. Before USB, every printer, mouse or keyboard needed its own connector. USB created a single standard. MCP does the same for the connections between language models and the outside world.

Current challenges

A standard on its way to winning

MCP adoption has been lightning fast: from a few hundred servers at launch in late 2024, the protocol now counts more than 10,000 active servers and 97 million monthly SDK downloads. OpenAI, Google DeepMind and Microsoft adopted it in 2025, and Anthropic handed the protocol over to the Agentic AI Foundation, a Linux Foundation fund created on December 9, 2025.

What it changes for a CDO

For a decision-maker, MCP means AI agents are no longer locked into one ecosystem. An agent built with Claude can use the same MCP servers as an agent built with GPT-4 or Gemini. This lowers vendor lock-in risk and speeds up deployment — instead of coding integrations, you connect existing servers.

Frequently Asked Questions

What is the Model Context Protocol (MCP)?

MCP is an open protocol created by Anthropic in November 2024 that standardizes how AI agents connect to external tools and data sources. In practice, it's a universal plug: instead of coding a specific integration for each tool (Gmail, Slack, Figma, a database), an MCP-compatible agent can connect to any MCP server through a common protocol.

What is the difference between MCP and a classic API?

A classic API is designed for human developers: documentation, endpoints, manual authentication. MCP is designed for AI agents: the server describes its capabilities in a format the language model understands directly. The agent can discover the available tools, understand how to use them and call them autonomously — without a developer having to code each integration.

Which tools support MCP?

In 2026, MCP is natively supported by Claude, ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code. On the server side, there are MCP connectors for Figma, GitHub, Slack, Google Drive, SQL databases and thousands of other tools. Since December 2025, the standard has been hosted by the Agentic AI Foundation, set up within the Linux Foundation by Anthropic, Block and OpenAI, with AWS, Bloomberg, Cloudflare, Google and Microsoft among its Platinum members.

Resources and Documentation