Anthropic open-sources the Model Context Protocol

Background

AI assistants are only as useful as the information and tools they can reach, such as files, chat channels, code repositories and databases. Anthropic said that before the Model Context Protocol, every new data source required its own custom implementation. A team that wanted an assistant to work with several systems had to build and maintain each connection separately. Without a shared standard, connecting M assistants to N tools means up to M times N separate connections, and each one must be maintained as either side changes.^[1]

What happened

On 25 November 2024 Anthropic open-sourced the Model Context Protocol (MCP), a new standard for connecting AI assistants to the systems where data lives. The aim was to let developers build against one standard protocol instead of writing a separate connector for each source.^[1]

Anthropic released three components together: the MCP specification and software development kits (SDKs), local MCP server support in the Claude Desktop apps, and an open-source repository of MCP servers.^[1]

Key details

MCP uses a client-server design. An AI application acts as the client, and each MCP server exposes a particular data source or tool through the same protocol, so any compatible client can use it. A server built once can therefore serve many assistants.

The specification defines how clients and servers exchange messages, and the SDKs give developers ready-made code to implement either side. Together they make the parts interchangeable: a client that follows the specification can use any server that does, whoever wrote it.

The repository included pre-built servers for Google Drive, Slack, GitHub, Git, Postgres and Puppeteer.^[1] Block and Apollo were early adopters, and the development-tool companies Zed, Replit, Codeium and Sourcegraph were working with MCP.^[1] Local server support meant the Claude Desktop apps could connect to servers running on a user's own computer, where many files and tools are kept.

Impact

On 9 December 2025 Anthropic donated MCP to the Agentic AI Foundation, a directed fund under the Linux Foundation. The fund was co-founded by Anthropic, Block and OpenAI, with support from Google, Microsoft, AWS, Cloudflare and Bloomberg.^[2] Open standards are often governed by foundations like this so that no single company controls their direction.

At the time of the donation, Anthropic reported more than 10,000 active public MCP servers and more than 97 million monthly SDK downloads across Python and TypeScript.^[2] Anthropic also said that products including ChatGPT, Cursor, Gemini, Microsoft Copilot and Visual Studio Code had adopted MCP, as had infrastructure providers AWS, Cloudflare, Google Cloud and Microsoft Azure.^[2]

Because competing companies adopted the same protocol, its value lies largely in the shared standard itself: a connection built once can work across many assistants and tools. The figures above are Anthropic's own reports, and they measure how widely the protocol is used rather than how well any particular integration performs.

Foundation governance matters most for a protocol that many competitors depend on, because changes to it are then made under shared, published rules rather than by the product team of a single company.