Building Applications with Built-in and Custom MCP in Studio

Mistral AI has launched Connectors in Studio to support the development of custom AI applications based on enterprise data. All built-in connectors and custom MCP (Model Context Protocol) are now immediately available via API and SDK when calling models and agents.

Developers can precisely control when tools are called through direct tool calling, and can implement human-in-the-loop approval flows to ensure security review and governance before tool execution.

Key features include the following:

• Programmatic control: Create, modify, and delete connectors, as well as list tools and execute them directly
• Centralized management: Registered connectors can be shared across all of Mistral's apps, including LeChat and AI Studio
• Support for complex workflows: Integrate with enterprise systems such as CRM and knowledge bases via the Conversation API, Completions API, and Agent SDK

Previously, building AI agents was inefficient because integration layers—checking API documentation, setting up OAuth, refreshing tokens, and so on—had to be written from scratch each time. Connectors packages the integration process into the MCP protocol as a reusable entity, providing an environment where tools can be executed anywhere with just one-time setup.