Here's what you did. You installed OpenClaw with npm install -g openclaw@latest, ran the onboarding wizard, and started the Gateway daemon as a system service. You connected at least one messaging platform and chose an AI model.
You wrote 4 config files: SOUL.md for personality, AGENTS.md for safety boundaries, IDENTITY.md for presentation, and USER.md for personal context. You learned how the permission system works in 3 layers (identity, scope, model) and how tool policies and sandbox mode add further protection.
In the next section, I'll show you how to give your assistant tools so it can do more than just chat. That's where things get interesting.