As a reminder, Anthropic is the firm behind the Claude 3 chatbot, a very serious competitor to the great ChatGPT. It was founded in 2021 by former OpenAI employeesnotably by Dario Amodei, who was previously vice president of research at OpenAI.
Nowadays, these virtual assistants face a limiting paradox: although they excel in analysis and reasoning (and even then, this sometimes fails), they struggle to exploit specific company datalocked in different software and databases. In response, Anthropic is launching the Model Context Protocol (MCP), a technical initiative that could redefine how AI interacts with information systems.
Bridges to connect AI to data
The current problem is concrete: each time a company wants to allow an AI assistant to access its internal documents, its business tools or its code bases, it must develop tailor-made connectors. This multiplication of specific developments hinders the adoption of large-scale AI models.
The MCP protocol proposes to unify these connections through an open technical standard, allowing developers to create the necessary bridges between their data and any AI assistant once. The MCP would act a bit like an instant translator capable of converting any programming language or data format into a language understandable by all chatbots. Thanks to this universal translatordata could circulate freely and be used optimally.
Another advantage, its open architectureas Anthropic explained: “ We are committed to developing MCP as a collaborative, open source project and ecosystem “.
Leading players already mobilized
The initiative already seems to be bearing fruit with adoption of the protocol by several reference companies. Block and Apollo have integrated MCP into their systems, while development tool companies like Replit, Codeium and Sourcegraph are actively preparing for its deployment. Anthropic also provides connectors Pre-designed for popular platforms like Google Drive, Slack and GitHub.
Below, a tweet from Alex Albert, head of customer relations for Claude, demoing the MCP protocol. The latter explains: “ Here’s a quick overview with the Claude desktop app: Using MCP, Claude can connect directly to GitHub, create a repository, and propose changes in just a few clicks. This integration only took less than an hour to set up after configuring the MCP “.
Here’s a quick demo using the Claude desktop app, where we’ve configured MCP:
Watch Claude connect directly to GitHub, create a new repo, and make a PR through a simple MCP integration.
Once MCP was set up in Claude desktop, building this integration took less than an hour. pic.twitter.com/xseX89Z2PD
— Alex Albert (@alexalbert__) November 25, 2024
The protocol allows developers to create data-exposing MCP servers and MCP clients – applications and workflows – connecting to it on demand. Subscribers to the Claude Enterprise plan can already connect the Claude chatbot to their internal systems via these servers.
In theory, this new technical approach is a solid idea: however, its success is not necessarily a guarantee. Obviously, Anthropic is not the only one to develop this type of solution and some of its competitors, such as OpenAI, are also developing their own proprietary data integration solutionslike the “ Work with Apps » by ChatGPT. Furthermore, the real effectiveness of the MCP protocol remains to be demonstrated, as Anthropic has not yet published comparative evaluations of its performance, simply communication videos.
However, standardizing data access for AI models constitutes an important step forward for the sector. Anthropic’s proposal could hit the mark, provided it convinces a sufficiently large ecosystem of partners and users.
- Anthropic launches Model Context Protocol (MCP) to simplify AI access to enterprise data via an open standard.
- MCP eliminates the need for custom connectors by acting as a universal translator between systems and AI assistants.
- The latter is already available and has been adopted by several companies.
