Usage Guide

Learn how to use OpenClaw effectively

Chat Commands

Send these in WhatsApp/Telegram/Slack/Google Chat/Microsoft Teams/WebChat (group commands are owner-only)

/status

Compact session status (model + tokens, cost when available)

/new or /reset

Reset the session

/compact

Compact session context (summary)

/think <level>

off|minimal|low|medium|high|xhigh (GPT-5.2 + Codex models only)

/verbose on|off

Toggle verbose output

/usage off|tokens|full

Per-response usage footer

/context list

Show loaded context (files, TRUNCATED if over limit)

/restart

Restart the gateway (owner-only in groups)

/activation mention|always

Group activation toggle (groups only)

Basic Interaction

Simply send a message to your connected channel (WhatsApp, Telegram, etc.) and OpenClaw will respond. The assistant understands natural language and can handle complex requests.

Tip: Be specific when you ask for something—"summarize my unread emails from the last 24 hours" works better than "check my email." For multi-step tasks, one message is enough; OpenClaw will break it down and use tools as needed.

Example Conversation

You: "What's on my calendar today?"
OpenClaw: [Checks your calendar and responds with today's events]

You: "Send an email to john@example.com about the meeting"
OpenClaw: [Composes and sends the email]

You: "Check the weather and remind me to bring an umbrella if it's raining"
OpenClaw: [Checks weather, sets reminder if needed]

Sessions and context

Each chat thread is a session. Use /new or /reset when you want a clean slate; use /compact when the thread is long but you want to keep going. /status shows the active model and token use; /usage tokens|full adds per-reply cost footers when your provider reports usage.

Loaded files and memory snippets appear in /context list. For how compaction and long-term memory work, see Context and Memory.

Tools and skills

OpenClaw can run shell commands, read files, use the browser, and call APIs when tools are enabled. If the agent replies but never acts, set a full tools profile in config—for example "tools": { "profile": "full" }. See FAQ: Agent doesn't run tools.

Skills extend what the agent can do (calendar, GitHub, smart home, and more). Install from ClawHub with openclaw skills install <name>; review untrusted skills first: Safe install checklist. List installed skills with openclaw skills list.

Groups and who can message you

In group chats, owner-only commands (like /restart) protect the gateway from random members. Use /activation mention|always to control whether the bot responds on every message or only when @mentioned.

For direct messages, configure pairing and allowlists before exposing a bot publicly. See Is OpenClaw safe? for a trust checklist.

WebChat and Control UI

You can chat in the browser via WebChat on the same machine as the Gateway, or through the macOS menu bar app. WebChat is useful for testing before linking WhatsApp or Telegram.

Automation from chat

Ask the agent to set reminders, run cron-style jobs, or react to webhooks—the gateway handles schedules and hooks. Start with Automation and the automation tutorial. Prefer a cheaper model for heartbeat checks; see Example setups & model routing.

Best practices

  • Be specific — Include dates, file paths, and expected output format.
  • One goal per message for complex work, or say “do this in order: …” so the agent plans steps.
  • Back up config before big changes: openclaw backup create (workspace backup).
  • Check logs when something fails: openclaw gateway status, then Troubleshooting.

Next steps