Let your Mac finish the job. Control it from your phone. Lock it when it's done.

No. The cloud can only request a fixed set of allow-listed actions — lock, sleep, keep-awake, volume, and so on — each identified by a strict enum, and the agent refuses anything outside it. There is no remote shell, from the iPhone app or anywhere else. (Locally-approved custom scripts are a planned extension that will keep the same approve-on-the-machine model.)

The macOS agent runs on macOS 12 Monterey and later as a lightweight menu-bar app, with lid-closed keep-awake on macOS 13 Ventura and later. A native iPhone app (coming soon) will support iOS 17 and later, and an Apple Watch app (coming soon) will support watchOS 10 and later. Windows support is on the roadmap; the data model and CLI are already platform-agnostic.

Coming soon. A native iPhone app for controlling your paired Macs, viewing activity, managing shortcuts, and triggering safe actions — including Siri/App Intents for voice-triggered actions like lock, sleep, lock + sleep, mute, volume up, volume down, and play/pause — is on the way to the App Store, alongside an Apple Watch app. Until they ship, the web dashboard works great from your phone's browser.

An Apple Watch app and a watch-face status complication are coming soon, alongside the iPhone app. The watch app shows each Mac's online status and lets you fire the same safe actions from your wrist; it pairs through your iPhone. Until then, use the web dashboard or CLI.

Only the macOS agent is required — it's what keeps your Mac awake and executes actions locally. You can trigger and cancel actions from anywhere with the web dashboard, CLI, webhooks, or Siri shortcuts today; the iPhone and Apple Watch apps will join them as additional control surfaces when they ship.

No. Like every CommandLatch control surface, the iPhone and Apple Watch apps will only request CommandLatch's fixed allow-listed actions. They will not be able to run shell commands, read files, see your screen, type, click, or remote-control the Mac.

Those give your phone a live window into your Mac — a remote shell, your screen, or an AI agent that can read files and run commands. CommandLatch is the opposite: your phone gets a fixed set of safe actions — lock, sleep, keep-awake, mute, volume — and nothing else. It’s also tool-agnostic: it keeps any long job alive (builds, CI, renders, uploads, agent runs) and latches the Mac down when it ends, not just one app’s sessions. The two are complementary — keep your Mac awake so a remote session survives the run, then lock + sleep it the moment work stops.

Yes. The web dashboard, CLI, webhooks, Siri shortcuts, and Claude Code hooks are all available today — the full workflow works without the iPhone app, and the iPhone and Apple Watch apps will be additional control surfaces once they ship.

Yes — the alert action sends a push notification to your iPhone (for example from the CLI: commandlatch alert "build done", or from a webhook or Siri shortcut). You can set up alerts today; delivery to your iPhone starts once the iPhone app ships and registers for notifications.

Yes. Give any paired Mac a friendly display name from the web dashboard; it shows everywhere — dashboard, CLI, and activity log — and (soon) in the menu-bar app and the iPhone and Watch apps. Clearing the name reverts to the hostname.

CommandLatch Pro is $6/month after a 7-day free trial (no card to start). One subscription covers every Mac on your account — it's per account, not per device. Manage or cancel anytime from the secure billing portal in the web dashboard. When a trial or subscription ends, your agent keeps running but remote commands pause until you resubscribe.

Commands expire 60 seconds after they’re created. If the agent doesn’t pick one up in time, it’s marked “expired” in your activity log and never runs — so a stale command can’t fire hours later.

No. Keep-awake only prevents sleep — it never disables your lock screen or password. When the session ends or you send a lock, your normal security applies immediately.

Commands route through your own account on a Supabase-backed backend with row-level security, so devices and logs are scoped to you. The desktop agent keeps its pairing credentials in a permission-restricted local file today, with OS keychain storage planned as a hardening step — and it forgets that token when you unpair or remove the device.