Anthropic's new open protocol lets AI systems tap into any data source
Key Points
- Anthropic has released an open-source protocol called Model Context Protocol (MCP) that provides a standard way for AI systems to access various data sources, including content repositories, business tools, and development environments.
- MCP uses a client-server architecture where developers can create MCP servers to expose their data sources or build AI applications (MCP clients) that connect to these servers. Anthropic has built MCP servers for platforms like Google Drive, Slack, GitHub, Git, Postgres, and Puppeteer.
- Currently, MCP only supports local connections, but Anthropic is developing support for remote servers with enterprise-grade authentication. The company emphasizes that MCP is fully open source and encourages developers to contribute to the protocol's development.
Anthropic has released an open-source protocol that lets AI systems connect to different kinds of data sources through a single standard interface.
Called the Model Context Protocol (MCP), it aims to simplify how AI assistants access content repositories, business tools, and development environments.
It works through a basic client-server setup. Developers can either create MCP servers that share their data sources or build AI applications that connect to these servers. To help get started, Anthropic has already built ready-to-use MCP servers for common platforms like Google Drive, Slack, GitHub, Git, Postgres and Puppeteer.
According to Anthropic, Block and Apollo have already integrated MCP into their systems, and several development toolmakers including Zed, Replit, Codeium and Sourcegraph are working on adding support.
"We're building a world where AI connects to any data source through a single, elegant protocol—MCP is the universal translator," writes Alex Albert, Head of Claude Relations at Anthropic.
MCP in the Claude desktop app connects to GitHub. | Video: Albert via X
Starting local
Currently, MCP only works with local connections - the servers need to run on your own computer. Instructions are available here.
But Anthropic says it's working on support for remote servers with enterprise-grade security, which will let teams share their data sources securely across organizations.
The company stresses that MCP is completely open source and wants developers to build their own MCP servers and help improve the protocol. For those interested in getting started, Anthropic offers software development kits for Python and TypeScript.
AI News Without the Hype – Curated by Humans
As a THE DECODER subscriber, you get ad-free reading, our weekly AI newsletter, the exclusive "AI Radar" Frontier Report 6× per year, access to comments, and our complete archive.
Subscribe now