Documentation Menu
Common Tasks
Frequently used console commands for managing your agent.
On this page
These are the tasks you will reach for most often when working in the console.
Check Agent Status
See if the agent process is running:
ps aux | grep node
A healthy agent shows one or more Node.js processes. If nothing appears, the agent may need to be restarted from the dashboard.
View Auth Profiles
Confirm which providers and keys are configured:
cat /home/node/.openclaw/agents/main/agent/auth-profiles.json | head -20
Look for entries like anthropic:default or openai:default. Each entry maps a provider and profile name to an API key.
Edit Configuration
Make changes to the agent config directly:
vi /home/node/.openclaw/openclaw.json
The config file uses JSON5 format (comments and trailing commas are allowed). Changes are hot-reloaded, so the agent picks them up without a full restart.
Check Disk Usage
Each agent has a persistent volume. Monitor how much space is used:
df -h /home/node/.openclaw/
du -sh /home/node/.openclaw/*
View Installed Skills
List what skills your agent has available:
ls /home/node/.openclaw/skills/
Skills are tools and capabilities that extend what your agent can do.
Diagnose Issues
Run the built-in diagnostic tool to identify and auto-repair common problems:
openclaw doctor --fix
This checks configuration files, API key validity, channel connections, and runtime health. The --fix flag attempts to repair any issues it finds automatically.
Check Channel Status
See which messaging channels are connected and their current state:
openclaw channels list
This shows all configured channels (WhatsApp, Signal, iMessage, email) and whether each one is active, disconnected, or errored.
Install Skills from ClawHub
Add new capabilities to your agent from the ClawHub skills registry:
npx clawhub install <skill-name>
For more on finding and installing skills, see Installing Skills.
Restart the Agent Process
If the agent is behaving unexpectedly, you can restart it from the console:
kill -SIGUSR1 $(pgrep -f openclaw)
This sends a reload signal to the OpenClaw process. For a harder restart, use the Stop and Start controls on the dashboard instead. That scales the entire container down and back up, giving you a clean slate.
Related Docs
- Running Commands — navigation and command basics
- Installing Skills — add new capabilities via the console
- Agent Settings — what you can configure in openclaw.json