CLI reference

clawmetry onboard / clawmetry setup#

The first-run wizard. setup is the same command framed as "connect to ClawMetry Cloud"; both accept the same flags and do the same work.

clawmetry onboard

What it does#

1. Detects the agent runtimes on this machine.
2. Asks whether you want a cloud account, or takes the answer from a flag.
3. Writes configuration into ~/.clawmetry/.
4. Starts the sync daemon and supervises it (launchd, systemd or the platform

equivalent).

1. Opens the dashboard.

Flags#

Local only#

clawmetry onboard --local

This writes a persistent local-only marker. A later clawmetry connect refuses to override it unless given --force, which exists so an automated provisioning run cannot quietly opt a machine into cloud sync that somebody deliberately opted out of.

Everything works in local mode: all 30 runtimes, every dashboard page, Guard, detectors, the MCP server, the query API. What you give up is the cloud dashboard, fleet view and phone approvals.

Non-interactive#

For provisioning:

# local-only node, no prompts at all
clawmetry onboard --local --node-id build-runner-07

# cloud node with a key you already hold
clawmetry onboard --key "$CLAWMETRY_KEY" --node-id build-runner-07 --cloud

Both forms are safe to run more than once — a second run reconciles rather than duplicating.

Node names#

--node-id defaults to the hostname, which is fine on a laptop and useless on a fleet of identically-named containers. Set it to something you would recognise in a list. Node identity is covered in Nodes and fleet.

Running the daemon yourself#

If you supervise processes with something of your own — supervisord, a systemd unit you manage, a container entrypoint — do not let the wizard install its own supervision:

clawmetry connect --key cm_xxx --no-daemon
# then run: python -m clawmetry.sync

See Run it as a service.

Troubleshooting the wizard#