Mistral Studio ships Managed Agents that run the same code locally, self-hosted, or managed
On November 5, Mistral’s Studio release notes added Managed Agents in public preview: build an agent once with the Agents SDK, then run it in-process, self-host with `agents serve`, or hand it to Mistral with `agents deploy`. Managed runs can work for hours or days in the background, resume after a crash, wait for human approval, and send traces into Studio by default. The same day also shipped a workflow builder and Apps CLI. My take: enterprises often stall not because the model is dumb, but because long jobs die and nobody trusts a restart. Lock approval gates and tool permissions first, then decide whether managed hosting is worth it.