Anthropic is proposing a new standard for connecting AI assistants to systems where data resides.
This standard is called Model Context Protocol, or MCP for short. Anthropic says that this standard, which is open source today, can help AI models generate better, more relevant query responses.
MCP Allows models - not limited to Anthropic's models - to pull data from sources such as commercial tools and software to accomplish tasks, as well as from content repositories and application development environments.
"As AI assistants become more widely adopted, the industry has invested significant resources in modeling capabilities, rapidly making inroads into inference and quality," Anthropic said in its Blog Posts writes, "However, even the most advanced models are limited by their isolation from data - trapped by information silos and legacy systems. Each new data source requires a custom implementation, making a truly connected system difficult to scale."
MCP ostensibly solves this problem through a protocol that enables developers to establish bi-directional connections between data sources and AI-driven applications such as chatbots. Developers can expose data through "MCP servers" and build "MCP clients" (e.g., applications and workflows) to connect to those servers on demand.
Anthropic says companies including Block and Apollo have integrated MCP into their systems, while Replit, Codeium and Sourcegraph Development tool companies, such as the MCP, are adding MCP support to their platforms.
"Developers can now build based on a standard protocol without having to maintain separate connectors for each data source," Anthropic wrote. "As the ecosystem matures, AI systems will maintain context as they move between tools and datasets, replacing today's fragmented approach to integration with a more sustainable architecture."
Developers can now start building with the MCP connector while subscribing to Anthropic's Claude Users of the Enterprise program can connect their company's Claude chatbot to their internal systems via an MCP server.Anthropic has shared pre-built MCP servers for enterprise systems such as Google Drive, Slack, and GitHub, and has said that a toolkit for deploying production-grade MCP servers to support the entire organization will soon be available. toolkit for deploying production-grade MCP servers that support entire organizations.
"We are committed to building MCP as a collaborative open source project and ecosystem," Anthropic wrote. "We invite [developers] to come together to build the future of context-aware AI."
In theory, MCP sounds like a good idea. But it's not clear that it will gain widespread support, especially among competitors like OpenAI, which may prefer that customers and ecosystem partners use the they (for inanimate objects) The data connection methods and specifications of the
In fact, OpenAI recently launched a new platform for its AI-powered chatbot ChatGPT Introduced a data connectivity feature that allows ChatGPT to read code in developer-facing code editing apps-similar to the use cases driven by MCP OpenAI says it plans to call this Work with Apps functionality to other types of applications, but it chose to implement it with a close partner rather than open source the underlying technology.
Moreover, it remains to be seen whether MCP is as beneficial and efficient as Anthropic claims. For example, the company says that MCP enables AI robots to "better retrieve relevant information to further understand the context associated with the coding task," but it has not provided any benchmarks to back up this claim.
