Documentation
Installation
Getting started with OpenClaw is simple. Follow these steps based on your operating system.
Requirements
- Node.js ≥ 22.x
- npm, pnpm, or bun
- Windows, macOS, or Linux
Global Installation
First-time Setup (Onboarding)
The onboarding wizard will guide you through initial setup:
This command will:
- Install and register the Gateway daemon
- Configure default workspace and models
- Set up OAuth subscriptions (Anthropic Claude / OpenAI GPT)
Starting the Gateway
Configuration
OpenClaw stores configuration in ~/.openclaw/openclaw.json
Example Configuration
Configuration Options
- agent.model: AI model to use (e.g., anthropic/claude-opus-4-5, openai/gpt-4)
- gateway.port: WebSocket server port (default: 18789)
- security.sandbox: Enable Docker sandbox for public channels
CLI Commands
Gateway Management
Send Messages
Direct Agent Interaction
Update OpenClaw
Diagnostics
Setting Up Integrations
Each platform requires specific configuration. Here are quick guides:
- Install the WhatsApp channel package
- Scan QR code to pair your phone
- Send a message to your OpenClaw number
Telegram
- Create a bot via @BotFather
- Add the bot token to your configuration
- Start chatting with your bot
Discord
- Create a Discord application
- Add bot to your server
- Configure bot token in OpenClaw
For detailed integration guides, visit the official documentation.
Frequently Asked Questions
Is my data sent to external servers?
No. OpenClaw runs entirely on your device. Only API calls to your chosen AI model provider (if using cloud models) are made.
Can I use local/open-source AI models?
Yes! OpenClaw supports local models via Ollama, LM Studio, and other OpenAI-compatible APIs.
How do I update OpenClaw?
Run openclaw update to update to the latest version. You can switch between stable, beta, and dev channels.
Can I run OpenClaw on a server?
Absolutely! OpenClaw works great on VPS, cloud instances, or home servers. Just ensure Node.js 22+ is installed.
Is OpenClaw free?
Yes, OpenClaw is open source under the MIT license. You only pay for AI model API usage if using commercial providers.