Running the gateway locally is the standard OpenClaw workflow. The main thing that trips people up is trying to debug channels or models before they have confirmed the gateway process is healthy.
Quick answer
Run onboarding in local mode, confirm openclaw gateway status --require-rpc, and only then open the dashboard or attach channels.
Command line steps
1. Write the local-mode config
Onboarding or setup should set gateway.mode=local. If it does not, repair that first.
BASH
$openclaw onboard --mode local
openclaw gateway status
2. Start or restart the gateway
If the service already exists, restart it instead of launching duplicate foreground processes.
BASH
$openclaw gateway restart
openclaw gateway status --require-rpc
3. Open the control surface only after the probe is green
A working dashboard is a confirmation step, not the first diagnostic step.
BASH
$openclaw dashboard
openclaw dashboard --no-open
What to check if it still fails
- If the gateway refuses to start because
gateway.modeis missing, treat that as config damage and runopenclaw doctor --fix. - If the RPC probe fails, read live logs with
openclaw logs --followbefore you touch model settings. - If you need a foreground debug session,
openclaw gateway --verboseis better than repeatedly restarting blind.