Anthropic launches Model Context Protocol to make it easier to connect AI assistants to data sources

Nov. 26 (Portaltic/EP) –

Anthropic has published the open source your Model Context Protocol (MCP for its acronym in English), which allows connect Artificial Intelligence (AI) assistants to the systems where data is storedfor example, content repositories, facilitating a universal connection and therefore helping models to offer better and more relevant answers.

Technology companies continue to work on developing the capabilities of their AI models, in order to achieve advances in reasoning and the quality of their responses. However, Anthropic has warned that “even the most sophisticated models” are limited, in part due to their “isolation from data,” that are relegated to storage systems that models do not always have access to.

This is because each new data sourcesuch as content repositories, business tools or development environments, require your own custom implementation so that models can access to its use, which “hinders the scalability of truly connected systems.”

In this framework, in order to offer a solution and allow AI assistants to connect to any data system, Anthropic has launched its Model Context Protocol, with which it aims to help the AI models produce more elaborate and relevant responsesto satisfy the needs of users.

Specifically, as the company has defined it in a communicated on his blog, it is a Universal, open standard developed to connect AI systems with data sources“replacing fragmented integrations with a single protocol”.

As detailed by Claude Anthropic’s director of relations, Alex Albert, in a post on (former Twitter), the MCP solves the “core challenge” of large language model (LLM) applications, which is “connecting them to their data.” Now, developers can integrate the MCP with their AI tool once and then connect this AI system to any “data source anywhere,” as it protocol shares “resources, tools and notices.”

Thus, as detailed by Anthropic, the architecture of this protocol is “simple”, since developers can expose their data through MCP servers or create AI applications that connect to these servers.

With all this, the company has already shared Pre-designed MCP servers for business systems such as Google Drive, Slack, GitHubPostgres and Puppeteer, among others. Likewise, some coding software, such as Replit, Codeium and Sourcegraph, are already using MCP to develop their AI agents.

Currently, this protocol is available to current Claude for Work clients, that they can begin testing the MCP servers locally, that is, by connecting Claude to internal data systems, as detailed by the technology company.

Finally, Anthropic has shared its intention to continue developing MCP as an “open source collaborative” project and ecosystem, with a view to facilitate its use for all developers and companies.