AI/TLDR

Mistral AI · 2026-04-27 · major

Mistral Workflows — Durable Agent Orchestration on Temporal, Now in Public Preview

Mistral ships Workflows in public preview: a Temporal-backed durable execution layer for production agents with state tracking, full audit trails, RBAC, and one-line human-in-the-loop checkpoints.

Mistral AI brand cover graphic

Mistral's enterprise stack gets a Temporal-based orchestration layer for production-grade agent processes.

Key specs

Stagepublic preview
Execution engineTemporal
Named customersASML, ABANCA, CMA-CGM, France Travail, La Banque Postale

What is it?

Workflows is the new orchestration layer inside Mistral's enterprise platform. It lets teams compose long-running, fault-tolerant business processes that combine Mistral models, agents, and connectors. It plugs into Studio for observability and pairs with hybrid deployment so workers can run on customer infrastructure while the control plane stays with Mistral.

How does it work?

Under the hood it's built on Temporal's durable execution engine, so each step is journaled and resumable: if a worker crashes mid-run the workflow picks up exactly where it left off. Single-line APIs add human-in-the-loop checkpoints. RBAC and workspace isolation cover enterprise security; CloudTrail-style audit lives in Studio.

Why does it matter?

The 'demo to production' gap is where most agent projects die — flaky LLM calls, partial failures, missing audit. Mistral is betting that durable execution plus first-party model integration is what unblocks that step. Named launch customers include ASML, La Banque Postale, France Travail, and CMA-CGM, signalling real enterprise uptake.

Sources

Tags

  • mistral
  • agents
  • workflows
  • temporal
  • durable-execution
  • enterprise
  • rbac
  • human-in-the-loop

← All releases · Learn AI