Manual

OpenClaw Usage Guide

Everything you need to know to use your OpenClaw AI agent — from first message to advanced commands.

Getting Started

After deploying your bot on OpenClaw Launch, you have three ways to start chatting:

Web Chat (easiest) — Click "Open Gateway" on your dashboard. This opens the gateway UI in your browser. No setup needed — just start typing.
Telegram — Find your bot on Telegram and send it a message. It will reply with a pairing code. Paste that code on your instance page to approve the connection. After pairing, the bot responds to your DMs.
Discord — Add your bot to a Discord server using the invite link from the configurator. DM the bot to get a pairing code, then approve it on your instance page. The bot will respond in DMs and mentioned channels.

Essential Commands

These are the commands you'll use most often. Type them directly in chat (web, Telegram, or Discord).

/new

Start a new conversation. Clears the current session context and begins fresh.

/stop

Stop the agent from generating. Useful when you want to interrupt a long response.

/help

Show available commands and usage information.

/commands

List all available slash commands with brief descriptions.

/status

Show current session info — active model, token usage, and session ID.

Model & Thinking

Control which AI model responds and how it reasons.

/model

View or change the AI model for the current session. Example: /model claude-sonnet-4.6

/think

Enable extended thinking mode. The agent will show its reasoning process before answering.

/reasoning

Toggle reasoning display on or off for the current session.

Session Management

Manage your conversation context and history.

/compact

Summarize and compress the conversation history to free up context space without losing key information.

/context

Show how much context window is used and how much remains.

/export

Export the current conversation as a text file.

Advanced

Power-user commands for direct system access.

/config

View or modify agent configuration for the current session.

/bash

Execute a shell command on the agent's container. Example: /bash ls -la

/exec

Run a code snippet directly. Supports multiple languages depending on installed skills.

Web Gateway

The web gateway is a full-featured UI that runs in your browser. Beyond chat, it provides:

Live config editor — Change your agent's system prompt, model, and skills without redeploying.
Session management — View active sessions, switch between conversations, and export chat history.
Device pairing — Approve Telegram/Discord pairing requests directly from the gateway UI.
Mobile support — The gateway is responsive and works on phones and tablets.

Pairing Flow (Telegram & Discord)

OpenClaw uses device pairing to keep your bot secure. Here's how it works:

Send any message to your bot on Telegram or Discord.
Copy the pairing code the bot sends back (a short alphanumeric code).
Paste it in the "DM Pairing" section on your instance page and click "Approve Pairing."
Done! The bot will now respond to your messages. Each user who wants to chat must pair separately.

Pairing is one-time per user. Once approved, the bot remembers the connection until the instance is redeployed.

Tips

Inline shortcuts

You don't always need slash commands. Just type naturally — the agent understands plain language. Use commands when you need specific system actions like switching models or starting a new session.

Session auto-reset

Sessions automatically reset after a period of inactivity (default: 30 minutes). Use /compact to preserve context in long conversations, or /new to start fresh immediately.

Switch models on the fly

You can change models mid-conversation. Use a fast model like Gemini Flash for quick questions, then switch to Claude Opus for complex reasoning — all in the same session.

Change the default model

The /model command only changes the model for the current session. To change the default model for all new sessions, use the model selector on your dashboard page.