Command line

Most people never touch a command line — you run Kiki from its desktop, the dashboard, or the mobile app. This page is for operators running headless servers and edge devices over SSH, the always-on machines Kiki was built for.

Updating the OS

Kiki updates as one sealed image with automatic rollback, so an update can never leave you with a half-broken system. On a device in the cloud, updates arrive over the air across your whole fleet. By hand:

sudo bootc switch ghcr.io/kiki-os/kiki-os-server:latest   # stage a new image
sudo bootc status                                          # show booted + staged
sudo systemctl reboot                                      # apply on next boot
sudo bootc rollback                                        # return to the previous image

Command	What it does
switch <ref>	Stage a new OS image to boot into.
status	Show the current and staged images.
rollback	Return to the previously booted image.

A failed boot rolls back on its own, so you shouldn't get stuck.

The agent service

The agent runs as a system service that starts on boot. You rarely touch it, but when you do:

sudo systemctl status agentd     # is it running?
sudo systemctl restart agentd    # restart, e.g. after editing config

Installing apps

You don't install apps from the command line. Browse the store in the dashboard or mobile app and install to any device — including headless ones — in a tap. Building and publishing your own app is done from your own machine with the kiki CLI, not on the device.

Adjusting the config file

A device reads its settings from /etc/kiki/agentd.toml. You almost never need to edit it — the privacy-preserving defaults are good, and a fleet is configured centrally from the dashboard. When operating a single headless box, the two settings worth knowing:

• Control mode — how much the agent does on its own (agent_mode is the default; also assisted_mode, human_mode, and bypass_permissions). See Who's in control.
• Cloud inference — off by default; turn it on to allow cloud models.

After editing, restart the service:

sudo systemctl restart agentd