Join our daily and weekly newsletters for the latest updates and exclusive content covering cutting-edge AI. Learn more
One of the decisions many companies must make when implementing AI use cases is connecting their data sources to the models they use.
Different frameworks like LangChain exist for integrating databases, but developers must write code every time they connect models to a new data source. Anthropic hopes to change this paradigm by releasing what it hopes will be a standard for data integration.
Anthropic published his Model Context Protocol (MCP) as an open source tool to provide users with a standard way to connect data sources to AI use cases. In a blog postAnthropic said the protocol would serve as a “universal and open standard” for connecting AI systems to data sources. The idea is that MCP allows models like Claude to directly query databases.
Alex Albert, Claude relations manager at Anthropic, said the that the company’s goal is to “build a world in which AI connects to any data source” with MCP as the “universal translator”.
“Part of the power of MCP is that it manages both local resources (your databases, files, services) and remote ones (APIs like Slack or GitHub) through the same protocol,” said Albert.
A standard way of integrating data sources not only makes it easier for developers to point large language models (LLMs) directly at information, but also eases data retrieval issues for companies building data agents. AI.
Since MCP is an open source project, the company said it encourages users to contribute to its project. deposit connectors and implementations.
A standard for data integration
There is not yet a standard way to connect data sources to models; this decision is left to enterprise users and model and database providers. Developers tend to write specific Python code or a LangChain instance to point LLMs to databases. Since each LLM works a little differently from one another, developers need separate code for each in order to connect to specific data sources. This often results in different models using the same databases without the ability to work together seamlessly.
Other companies are expanding their databases to make it easier to create vector embeddings that can connect to LLMs. One such example is Microsoft integrating its Azure SQL with Fabric. Smaller companies like Fastn also offer a different method for connecting data sources.
Anthropic, however, wants MCP to work even beyond Claude, as a step toward interoperability of models and data sources.
“MCP is an open standard that allows developers to establish secure, two-way connections between their data sources and AI-powered tools. The architecture is simple: developers can either expose their data through MCP servers or create AI applications (MCP clients) that connect to these servers,” Anthropic said in the blog.
Several social media commenters welcomed the MCP announcement, particularly the open source versions of the protocol. Some users in forums like Pirate News were more cautious, questioning the value of a standard like MCP.
Of course, MCP is currently only a standard for the Claude model family. However, Anthropic has released pre-built MCP servers for Google Drive, Slack, GitHub, Git, Postgres and Puppeteer.
VentureBeat has contacted Anthropic for additional comment.
The company said early adopters of MCP include Block and Apollo, with vendors like Zed, Replit, Sourcegraph and Codeium working on AI agents that use MCP to gain insights from data sources.
All developers interested in MCP can access the protocol immediately after installing the pre-built MCP servers via the Claude desktop application. Businesses can also build their own MCP server using Python or TypeScript.