Anthropic open-sourced a new protocol to connect data hubs to artificial intelligence (AI) systems on Monday. Dubbed Model Context Protocol (MCP), the company claimed it can overcome the limitations of traditional data integration methods and solve the problem of data silos. The AI firm is also offering local MCP support in the Claude Desktop apps and an open-source repository of MCP servers. Notably, the company is also offering pre-built MCP servers for Google Drive, Slack, GitHub, Git, Puppeteer and more.
Anthropic Open Sources Model Context Protocol
While large language models (LLMs) are pre-trained on massive datasets, they are often not enough, especially when an AI chatbot has to perform specific tasks. Additionally, the capability to upload files and folders to AI systems to get contextually aware responses about them has become a critical functionality of these tools.
However, when it comes to interacting with external datasets and knowledge hubs, AI models face several challenges. On a macro level, this mainly arises as every different external data source has unique ways it lets the AI scrape the information and process it. On a deeper level, the problem also arises due to the lack of a single protocol that AI developers can follow to access said data sources.
As a result, each AI system behaves differently when interacting with different external knowledge hubs and the success of outputs can vary vastly. In a blog post, Anthropic shared its Model Context Protocol (MCP) which can solve this problem. The company said MCP is a universal, open standard for connecting AI systems with data sources and replaces fragmented integrations with a single protocol.
The biggest benefit of this is a reliable way to provide AI systems access to the data they require, the company highlighted. The company has open-sourced three components of MCP for developers — MCP specifications and software development kits (SDKs), local MCP servers for Claude Desktop apps, and a repository of MCP servers.
Additionally, the AI firm also shared pre-built MCP servers for popular enterprise systems such as Google Drive, Slack, GitHub, Git, Postgres, and Puppeteer. Anthropic stated that companies such as Block and Apollo have already integrated MCP into their systems while development tool firms such as Zed, Replit, Codeium, and others are using MCP to improve their platforms.
Anthropic said that it will soon provide developer toolkits to deploy remote production MCP servers that can help enterprises connect AI systems to their organisation’s data hubs.