Mistral Studio Managed Agents: What the Update Adds

Mistral AI shipped a big upgrade to Mistral Studio on November 5, and it is aimed squarely at the hardest part of enterprise AI: getting a promising pilot into daily production. The update adds managed agents and apps to Studio, plus new tools for workflows, shared enterprise context, observability and evaluations. It pairs the platform with Mistral Large 4, the open-weight model the company nicknames Le Chonk.

• Mistral Studio now offers managed agents with skills, code sandboxes, subagents and hooks
• Workflows gain event triggers from enterprise systems and a visual graph view
• Apps get hosting, authentication and access controls, deployable with one command
• Studio runs in Mistral's cloud, in a customer VPC, on-premises, or a mix of all three

What Are Mistral Studio Managed Agents?

Managed agents handle complex, multi-step tasks while Studio runs the infrastructure underneath. Developers can give an agent skills that capture a team's procedures, sandboxes where it can run code, and subagents for parts of a job. Hooks add custom logic at key points in execution. Mistral says the runtime is the same one behind millions of Mistral Vibe sessions each week.

Mistral shared an internal example. Its engineers built an agent that starts investigating when a production alert fires. It follows the team's runbook, asks for approval before sensitive actions, and posts findings in Slack. According to Mistral, that saves the team hundreds of hours a month.

A new Mistral MCP server also lets developers build with Mistral Vibe or another coding agent, test locally, and deploy to Studio with one command.

How Do Studio Workflows and Apps Fit Together?

Mistral draws a clear line between the three building blocks. An agent investigates and acts. A workflow coordinates steps and approvals. An app gives employees a way to use both.

Workflows, which Mistral's docs describe as a public preview, are written in code and built to survive crashes and restarts, with durable state, retries and scheduling handled by the platform. Interrupted runs pick up where they left off. Teams can now describe a process in Mistral Vibe and have it generate, test and deploy a workflow with editable code. One internal workflow sorts product feedback into bugs and feature requests, which Mistral says saves its product operations team about 35 hours a week.

Apps get enterprise hosting, sign-in and access rules out of the box, and can draw on agents, workflows and document search indexes.

Why Deployment Control Matters for Enterprise AI

The theme running through the release is control. Customers can deploy both the models and Studio itself inside their own perimeter. They choose the models, govern which data and tools agents can reach, and decide which actions need a human sign-off. OpenTelemetry traces show every model call, tool use and approval in a run, and evaluations let teams test prompt or model changes against their own datasets before release.

Mistral names ABANCA, Accenture, Alten, BNP Paribas, Capgemini, CMA CGM and Tata Consultancy Services as customers and partners already using Studio in production. For the model side of the story, see our guide to running Mistral Large 4 open weights locally, and our look at Voxtral 3 voice agents. More platform news is in our AI section.

Sources: Mistral AI — November 5, 2026; Mistral Workflows documentation — accessed November 6, 2026.