# Mistral Studio Adds Managed Agents and Apps, With the Agent Runtime Still in Public Preview

Times of AI Desk · 2026-11-07 · Enterprise

[https://timesof.ai/2026/11/mistral-studio-managed-agents-apps](https://timesof.ai/2026/11/mistral-studio-managed-agents-apps)

> Mistral says it is expanding Studio with managed agents — including skills, sandboxes, subagents, hooks, approvals and tracing — and with enterprise apps that provide hosting, authentication and access controls, plus workflow upgrades such as event triggers, a visual graph and resumable runs. The pitch is paired with Mistral Large 4, also called Le Chonk. Documentation for Managed Agents states that the feature and the Agents SDK are in public preview. Mistral says Studio can run in the cloud, a customer's VPC or on-premises.

Mistral is packaging the model it calls **Le Chonk** as something enterprises can run as agents, workflows and internal apps under their own roof. In a product post, the company said it is expanding **Mistral Studio** with **managed agents** and **apps**, alongside upgrades to workflows. That sits on top of [Mistral Large 4, the trillion-parameter-class open-weight model it previewed last month](https://timesof.ai/2026/10/mistral-large-4-preview-1t-open-weights), and follows [Studio's public-preview Responses API](https://timesof.ai/2026/10/mistral-responses-api-public-preview).

Managed agents, as Mistral describes them, can be given skills, sandboxes, subagents, hooks and human approvals, with tracing on each step. Workflows gain event triggers from enterprise systems, a visual graph and runs that can resume after an interruption. Apps get hosting, authentication and access controls, with a path to test locally and deploy with one command. A Mistral MCP server is meant to let developers build in Mistral Vibe or another coding agent and deploy into Studio.

## Preview, not a quiet GA

Mistral's documentation is more cautious than the news post's launch language. The Managed Agents overview states that **Managed Agents and the Agents SDK are in public preview**, and that their API, behavior and limits can change in ways that may require code changes. Production deployments can be self-hosted (`agents serve`) or Mistral-managed (`agents deploy`). By default, the docs say, a session ends after 24 hours, or after four hours idle; both limits can be raised.

The company says Studio — models and platform — can run in the cloud, a customer's VPC or on-premises, or in a mix of those environments, and that customers retain control of data, tool access and which actions need approval. Customer names in the post include **ABANCA**, **BNP Paribas**, **Capgemini**, **CMA CGM** and **Tata Consultancy Services**, among others; the quotes describe existing Studio use rather than independently audited results from this release.

## Limits

- Managed Agents are in public preview per Mistral's docs, even though the news post describes the Studio expansion in present-tense shipping language.
- We did not find a similarly explicit "public preview" label on every Apps feature in the materials we read; treat app hosting claims as Mistral's product description.
- Customer testimonials are selected by Mistral and do not quantify this release.
- Timing for private-cloud or on-premises managed deployments beyond "wherever you choose" is not specified as a dated generally-available milestone in the news post we read.

## Sources

- [Mistral: Build Enterprise Agents, Workflows & Apps in Mistral Studio](https://mistral.ai/news/build-with-studio/)
- [Mistral Docs: Managed Agents overview](https://docs.mistral.ai/studio/agents/overview)