Is Mistral Studio's managed agents feature generally available?

Mistral Studio's managed agents, the Agents SDK, the workflow builder, the apps CLI and the evaluation tools are in public preview as of November 5, 2026. Mistral warns that the API, behaviour and limits can change during the preview, so an update may need code changes. Studio apps inside Vibe Code Web are the one piece marked generally available.

Can I run Mistral managed agents on my own servers?

Mistral managed agents run in three places. You can run an agent inside your own process, host it yourself with the `agents serve` command, or let Mistral run it with `agents deploy`. Agents deployed to Mistral run in the background and resume after a crash. The Studio post also lists cloud, VPC and on-premises deployment options for the platform.

How does the Mistral Studio workflow builder work?

The Mistral Studio workflow builder takes a plain-language description of an automation. Mistral Vibe then writes, tests and deploys the workflow, and you can still read and edit the generated code. Workflows mix fixed-rule steps, agent reasoning and human decisions, support resumable runs, and can start from webhooks and events in other enterprise tools.

How do I test and monitor agents built in Mistral Studio?

Mistral Studio now sends agent, model and tool spans to Studio by default, as OpenTelemetry traces with redaction on the worker. The Evaluation SDK, mistralai-evaluations, runs offline evaluations on managed datasets, and online evaluations score live production traces. Teams can also turn trace spans into dataset records in the Trace Explorer.

Who already uses Mistral Studio in production?

Mistral says customers and partners already run Mistral Studio in production, including ABANCA, Accenture, Alten, BNP Paribas, Capgemini, CMA CGM and Tata Consultancy Services. Mistral also uses it internally: one workflow saves its product operations team about 35 hours a week, and an incident-investigation agent saves hundreds of hours each month.

Mistral AI · 2026-11-05 · major

Mistral Studio managed agents — agents, apps and workflows ship to Studio

In short

Mistral Studio adds managed agents, a plain-language workflow builder, hosted apps, managed search indexes and offline and online evaluations, mostly in public preview, all built around Mistral Large 4.

Mistral AI · 2026-11-05 · 2 sources checked

Mistral Studio announcement banner for managed agents, apps and workflows

Mistral turns Studio into a place to build, deploy, trace and test agents and apps on Mistral Large 4.

Quick facts

Maker	Mistral AI
Product	Mistral Studio (console.mistral.ai)
What's new	Managed agents, workflow builder, apps CLI, evals, managed datasets, search index
Status	Mostly public preview; Studio apps in Vibe Code Web are GA
Agents SDK	mistralai-agents (Python)
Deployment	Mistral-hosted, cloud, VPC or on-premises

What is it?

Managed agents are the headline of the November 5 Mistral Studio release: multi-step agents with sandboxes, subagents, skills, tools and MCP connectors whose runtime Studio manages. The same release adds a workflow builder, hosted apps with login and access control, a managed vector search index, managed datasets, and offline and online evaluations. Mistral pitches it with one line: "Le Chonk brings the brains. Studio puts it to work."

How does it work?

Developers write agents in Python with the mistralai-agents SDK, test them in the terminal with `agents chat`, then ship with `agents serve` (self-hosted) or `agents deploy` (Mistral-hosted, durable, resumes after crashes). A sandbox gives the agent its own computer for shell commands and file edits. The new `mistral apps` CLI scaffolds and deploys Studio apps, and a Mistral MCP server lets an existing coding agent build and deploy to Studio with one command.

Why does it matter?

Enterprise teams on Mistral no longer need separate tools for agent hosting, tracing and evaluation, because traces, datasets and scores now live next to the agents themselves. Companies that must keep data in-house can run the same stack in a VPC or on-premises instead of on Mistral's cloud.

Who is it for?

Enterprise developers building agents on Mistral models

Frequently asked questions

Is Mistral Studio's managed agents feature generally available?: Mistral Studio's managed agents, the Agents SDK, the workflow builder, the apps CLI and the evaluation tools are in public preview as of November 5, 2026. Mistral warns that the API, behaviour and limits can change during the preview, so an update may need code changes. Studio apps inside Vibe Code Web are the one piece marked generally available.
Can I run Mistral managed agents on my own servers?: Mistral managed agents run in three places. You can run an agent inside your own process, host it yourself with the `agents serve` command, or let Mistral run it with `agents deploy`. Agents deployed to Mistral run in the background and resume after a crash. The Studio post also lists cloud, VPC and on-premises deployment options for the platform.
How does the Mistral Studio workflow builder work?: The Mistral Studio workflow builder takes a plain-language description of an automation. Mistral Vibe then writes, tests and deploys the workflow, and you can still read and edit the generated code. Workflows mix fixed-rule steps, agent reasoning and human decisions, support resumable runs, and can start from webhooks and events in other enterprise tools.
How do I test and monitor agents built in Mistral Studio?: Mistral Studio now sends agent, model and tool spans to Studio by default, as OpenTelemetry traces with redaction on the worker. The Evaluation SDK, mistralai-evaluations, runs offline evaluations on managed datasets, and online evaluations score live production traces. Teams can also turn trace spans into dataset records in the Trace Explorer.
Who already uses Mistral Studio in production?: Mistral says customers and partners already run Mistral Studio in production, including ABANCA, Accenture, Alten, BNP Paribas, Capgemini, CMA CGM and Tata Consultancy Services. Mistral also uses it internally: one workflow saves its product operations team about 35 hours a week, and an incident-investigation agent saves hundreds of hours each month.

Try it

Set MISTRAL_API_KEY, install mistralai-agents, then run `agents chat` (see docs.mistral.ai/studio/agents/overview)