Anthropic deploys MCP: a new era for AI connected to applications

Anthropic unveils Model Context Protocol (MCP), an open source standard that revolutionizes the connection between AI and professional applications. This standardized protocol facilitates the access of models like Claude to external data, paving the way for intelligent assistants truly integrated into workflows.

A persistent problem: isolation of AI

Language models like Claude, Chatgpt or Gemini shine in conversation, but suffer from a major problem: they remain isolated from business tools. Whether for code management, database consultation or interaction with tools like Slack, integrations remain specific and complex to maintain. This fragmentation limits the capacities of AI and slows down their adoption in the professional world.

With MCP, Anthropic offers a universal solution that simplifies the connection between AI and existing applications.

MCP: an open protocol to connect Claude to applications

MCP is an open protocol that allows developers to create standard connectors between AI models and various digital services. Concretely, MCP is based on a customer-server architecture:

• MCP servers: they exhibit data sources or specific features (e.g. access to Google Drive files, management of Slack messages, consultation of postgre databases).
• MCP customers: AI applications like Claude Desktop can connect to the context and carry out actions.

The objective is to simplify the integration of AI into professional tools without requiring complex personalized integrations.

MCP flower: an interface to democratize connections

If MCP remains a powerful technology, its technical implementation can put off non-development. This is where Fleur MCP comes in, an application developed by an independent team.

Fleur works like an MCP connectors App Store: instead of manually code integrations, users can install connectors with a click. Some examples available:

• Github: Consultation of deposits, creation of tickets, code comments;
• Browser: automated web navigation, form filling;
• Slack: recovery and summary of conversations;
• Obsidian: Research and management of personal notes.

Fleur is currently only available on MacOS, but its potential could accelerate the adoption of MCP.

Concrete and promising uses

Companies like Block or Apollo have already adopted MCP to connect their internal systems to Claude. Among the most promising use cases:

• Software development: generation and code analysis with Github.
• Automation of tasks: Direct interaction with project management tools like Linear.
• Information search: automated web navigation and relevant data recovery.

Thanks to MCP, AI models are no longer simple dialogue boxes. They can act, interact with applications and automate workflows.

Towards a more autonomous and integrated AI

With MCP, Anthropic adopts a modular approach, similar to Unix: instead of developing an AI that can do everything, the objective is to connect the AI ​​to the right tools.

One can imagine that in the future, Claude can:

• Reserve an Uber.
• Manage a professional agenda.
• Supervise a complete project on Github and Slack.

This evolution could well mark a major turning point in the use of conversational AIs. Thanks to MCP and Fleur, Claude becomes more than an assistant: he turns into a real digital operator.