Security and the power menu
Who can use the device, what the agent may do, and the controls that lock, restart, or shut it down.
The power menu
The power button at the end of the dock works the way a computer's power menu does. While the system is running it offers Lock screen, Restart, and Shut down; when the system is off, it offers Start instead — the options change with the state, so the right ones are always in front of you. Restart and Shut down ask you to confirm first.
Shut down stops the agent runtime — the part that thinks, runs tools, and talks to channels. The dashboard itself stays available, showing the red Offline dot and keeping the Start option within reach, the same way a sleeping machine still answers its own power button. The Online/Offline dot in the header reflects the runtime in real time.
The lock screen
Lock screen (in the same menu) covers the display immediately. Unlocking needs the password you set during setup. Locking does not sign you out, end the session, or stop the agent — anything running keeps running. It only stops someone walking past from using it.
Approval modes
Separately, you control what the agent may do without asking. This is the Safety choice from setup, changeable any time from the shield button beside the chat's send button — or in Config.
| Mode | Behaviour |
|---|---|
| Ask every time | Nothing that changes your machine happens without a prompt. Safest, most interruptions. |
| Smart | Reads freely; asks before anything destructive. The usual choice. |
| Trusted | Acts without asking. Fastest, and only sensible once you trust the setup. |
Where credentials live
API keys and channel tokens are stored on the device in an environment file, visible under Keys. They are never sent anywhere except to the provider they belong to.