Close the lid. The work keeps going.

Keep-running: one setting, off by default

Without the daemon, sessions close with the app — a deliberate default, because holding your work alive should be a choice. Turn keep-running on and quitting the window stops quitting the work: agents keep running, PTYs stay attached, automation flows keep firing.

Reopen the app and everything reattaches — nothing restarted, no context lost, scrollback intact.

The muster CLI

The running app answers to a CLI: list and status to see what’s in flight, wait to block on a task, spawn to create AND start one, attach to drop into a live session, close, reopen and resume to drive the lifecycle, events to stream what’s happening, policy to inspect and apply permission policies, and queue to list, enqueue, cancel, remove, retry and work the task queue. Scripts get the same control surface you do.

Work that outlives the command

muster queue run spawns into the daemon, so a queued agent keeps going after the command exits — fire it from a CI job or a cron line and walk away. It starts entries for SSH-remote projects too, bootstrapping and tunneling to the host exactly as the app does.

reload-config, honestly

muster reload-config live-applies the settings that are safe to change under a running system: post-create hooks, the default shell, and the automations master switch. Every other key is restart-only, and the CLI says so rather than pretending the reload took.

Per-user by construction

The control socket is per-user. Another account on the same machine can’t see your sessions, your queue or your agents — the boundary is the operating system’s, not a password field.