Keep your operator online (always-on & reliability)

An operator only earns while it’s serving. A session you started by hand stops when you close the terminal or reboot — so for real earnings, run it as a persistent service that starts on boot and restarts on failure.

Already serving? This is the last step after run an operator.

Why uptime matters

You only earn while online. Every minute offline is requests you don’t win.
Reputation compounds. Consistent, honest serving builds standing, and the market favours operators it can rely on — so uptime pays twice.

Install it as a service

The halo CLI wraps your platform’s service manager so you don’t have to write unit files:

halo service install serve   # install + start on boot
halo service status serve    # is it running?
halo service logs serve      # tail the logs

halo serve connects outbound to the relay over WebSocket — there’s no public URL or inbound port to expose or keep open — so a home machine behind a router works fine.

Watch and recover

Check health any time with halo doctor --json (node version, install + wallet state, provider, relay health).
Read the logs with halo service logs serve if requests stop arriving.
The service restarts automatically on crash or reboot, so a power blip or update won’t quietly take you offline.

Keep an eye on it from the web

Pair your operator wallet with a dashboard account to watch uptime, requests, and earnings from the browser — and to join the League: pair with the dashboard.

Practical tips

Put the machine on a stable connection and power (a UPS for a home rig).
If you serve a local model, make sure its runtime (Ollama / LM Studio) also starts on boot — otherwise the operator has nothing to serve. See serve local models on your GPU.
Prefer a box you don’t reboot often, or schedule updates for low-traffic hours.