Connect the dots: Build with built-in and custom MCPs in Studio
Mistral AI announced the release of Connectors in Studio, now in Public Preview, to support developers building AI applications grounded in enterprise data. According to the company, all built-in connectors and custom MCPs are available via API and SDK for use with all model and agent calls.
The release introduces direct tool calling, which Mistral describes as giving developers control over how and when tools are invoked without authentication barriers affecting testing and iteration. It also adds human-in-the-loop approval flows, allowing review and confirmation before tool execution.
Connectors can be created, modified, listed, and deleted programmatically, and their tools can be listed and run directly, Mistral said. All connectors are centrally registered and available across Mistral apps: LeChat and AI Studio, with Vibe coming soon. Usage via the Conversation API, Completions API, and Agent SDK is intended to support workflows and integration with enterprise systems such as CRMs, knowledgebases, and productivity tools.
Mistral said a connector packages an integration into a single reusable entity using the MCP protocol. Once registered, Mistral states a custom MCP connector is discoverable, governed, and monitored in Studio and becomes a native tool for any conversation, agent, or workflow. The company cited enterprise integration challenges including tracking API documentation, writing and maintaining tool functions, setting up OAuth, handling token refresh, and debugging pagination issues.
The announcement includes a walkthrough for a code repository auditing agent using GitHub, web search, and a custom DeepWiki MCP server registered through the connectors API. Mistral also demonstrated calling a connector tool directly, and a configuration using requires_confirmation to pause execution for explicit approval before a tool such as gmail_search runs.
Mistral said Connectors in Studio are available starting today via the Studio console, with documentation and cookbooks covering common usage patterns.
Based on reporting from the original publisher. Visit the source for full context and later updates.
Publisher excerpt
Connect enterprise data to your AI applications with reusable connectors, direct tool calling, and human-in-the-loop approval controls.