How Anthropic’s MCP might finally make AI less dumb about context

Anthropic has introduced a new open-source protocol for connecting AI systems to diverse data sources, potentially enhancing the efficiency of AI assistants. The Model Context Protocol (MCP), launched on November 25, 2024, aims to streamline interactions between AI applications and various data repositories.

Anthropic asserts that MCP resolves the challenge of isolated AI models which struggle to access data due to existing information silos. The protocol facilitates two-way communication between data sources and AI applications, allowing developers to create MCP servers and clients that interact seamlessly. This addresses inefficiencies in maintaining individual data connectors for each source, simplifying the integration process significantly.

In a demonstration using the Claude desktop application Alex Albert, Anthropic’s head of Claude relations illustrated how easy it is to integrate MCP for tasks such as connecting to GitHub and managing repository actions.

I just connected Claude to an internet search engine using MCP.

Here’s how you can do it too in under 5 minutes: pic.twitter.com/IT4j4AidXw

— Alex Albert (@alexalbert__) November 25, 2024

The setup for this connection reportedly took less than an hour. Notably, organizations like Block and Apollo, along with development tools such as Replit, Codeium, and Sourcegraph, have already begun implementing MCP into their frameworks. This suggests a growing adoption of the protocol among industry players.

Anthropic’s approach contrasts with efforts from competitors, particularly OpenAI, which recently unveiled its “Work with Apps” feature in ChatGPT. This functionality permits the AI assistant to interact with specific coding applications available on Mac. OpenAI’s solution appears tailored to selected partners, diverging from Anthropic’s broader protocol aimed at universal application across different tools and datasets.

MCP operates under a framework that encourages developers to build against a standardized protocol, which is expected to foster scalability and maintain context as AI systems evolve within various environments. Anthropic emphasizes that, as the ecosystem matures, AI systems can move fluidly between different resources without facing the fragmented integrations currently common in the industry. This vision aligns with a shift toward more integrated and context-aware AI applications.

Despite the promise of MCP, its practical efficacy remains to be seen. While Anthropic claims the protocol will enable AI models to retrieve contextually relevant data more effectively during coding tasks, it has not provided empirical benchmarks to substantiate these assertions. Industry observers are left questioning how well MCP will perform compared to other established frameworks, especially given the competitive landscape characterized by various proprietary models.

Developers are encouraged to start utilizing MCP connectors, particularly those subscribing to Anthropic’s Claude Enterprise plan, which allows for direct connections between Claude and internal data systems. Anthropic aims to support this initiative by offering prebuilt MCP servers compatible with major enterprise software like Google Drive, Slack, and GitHub. Plans are also in place to release toolkits that facilitate the deployment of production-ready MCP servers across organizations.

Featured image credit: Anthropic