MCP in Mistral AI's Vibe

Updated November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat is now Vibe (still at chat.mistral.ai): a single agent for work and code, carrying over your conversations, settings, and subscriptions. The legacy Le Chat experience remains accessible in Chat mode; in Work mode, agents are now skills (invoked with @), memories form the knowledge base, and deep search is the Deep Research skill. Plans are now called Mistral Free, Pro, Team, and Enterprise. This lesson uses the name Vibe; some menus described here may have changed names or locations.

Vibe: Your AI Assistant Augmented by MCP

Vibe, Mistral AI’s conversational assistant, supports the MCP protocol natively. You plug your professional tools and services directly into the conversation interface—no custom development required. Vibe shifts from a simple conversational agent to an automation platform that integrates with your entire work environment.

As of April 2026, it offered a rich directory of pre-integrated connectors, plus the ability to add your own MCP servers. This dual approach—an official catalog on one side, total openness on the other—makes it one of the most comprehensive MCP clients on the market.

The Connector Directory

The Connectors tab brings together two families of integrations. The first groups featured connectors, developed or validated by the Mistral AI team, guaranteeing a high level of reliability and security. You’ll find GitHub for repositories, issues, and pull requests; Notion for search and page creation; PayPal for transaction lookup; Hugging Face for models, Spaces, and datasets from the AI community; Linear for project management; Gmail for reading and managing emails via OAuth; Atlassian for Jira and Confluence; and Box for files stored in the cloud.

The second family consists of custom connectors. Vibe accepts any remote MCP server: provide its URL, and it becomes usable like the rest. This opens the door to your internal APIs, niche services no one has integrated yet, or tools published by the community.

Activating and Managing Connectors

To activate a connector from the directory, follow these four steps:

1. Open the Connectors tab in Vibe
2. Browse featured connectors or search for the one you want
3. Click Connect to start the connection process
4. Depending on the authentication type, provide a token, authorize OAuth access, or simply log in

The connector then appears with an active icon in your toolbar. You can disable it temporarily without deleting it, avoiding the need to reconfigure everything every time you switch topics.

This isn’t just a cosmetic detail—connectors are managed per conversation. Working on a GitHub project? Activate only GitHub for that session. Switching to sorting emails? Turn off GitHub and activate Gmail. The reasoning matches any agent system: the model selects the right tool more effectively when it has fewer options visible. Having a dozen connectors active simultaneously blurs selection and leads to approximate calls.

In an organization, administrators have additional controls. They can enable or block certain connectors for the entire company and maintain audit visibility into what teams are using. However, authentication remains individual: each member’s GitHub token is personal and grants access only to their own repositories, preventing a shared connection from becoming a governance vulnerability.

The Interface in Action

When a connector is active, a small tool icon appears in the input area. Hovering over it shows the list of tools available in the current conversation—a useful habit before formulating a complex request.

As soon as the model decides to use a tool, the interface shows the tool name, the arguments sent to the server in JSON, and two buttons: Allow to authorize execution, Decline to refuse it. This human validation—the famous “human in the loop”—is the primary security layer: nothing goes to your systems without your approval. An Always Allow option exists for tools you fully trust, but reserve it for read actions. Enabling it on a tool that writes to a shared repository removes the only safeguard you have.

Additional Features and Mobility

MCP doesn’t work alone in Vibe. Canvas mode previews HTML, CSS, or Markdown code and displays a visual result, such as a Google Maps iframe or a web page generated on the fly. Memories, launched alongside MCP support, let Vibe retain information about you to personalize responses, including importing preferences from other assistants. OCR and speech-to-text are integrated natively: no need to plug in an MCP server to read text in an image or transcribe a voice message.

Finally, all connectors available on the web also work in the mobile app. Checking a Jira ticket on the go, reviewing a Notion page before a meeting, or triggering a workflow from your phone works exactly the same as on desktop.

Key Takeaways

• Vibe natively integrates the MCP protocol with a directory of pre-integrated connectors
• You can add custom connectors with a simple URL
• Connectors are managed per conversation to optimize model performance
• Human validation at each tool call guarantees security
• Organization administrators control connectors available to their team
• MCP also works on Vibe’s mobile app