Getting Started with OpenClaw

I walked through this setup myself when I first installed OpenClaw—here's what actually works. You'll go from zero to a running assistant in about 15–20 minutes. We'll use the one-liner installer and the onboard wizard; you can tweak things later.

Command not found after install: Restart your terminal (or open a new tab). The installer adds OpenClaw to your PATH, but existing sessions don't see it until you reload.

Wizard hangs or QR code expires: Channel linking (Telegram, WhatsApp) can time out. Just run openclaw onboard again and pick the same channel; you don't have to redo everything.

Gateway won't start: Run openclaw doctor --fix. If you see port or token errors, check the Troubleshooting guide—most issues are fixable in a few steps.

Alternatively, use the Setup Wizard in your browser to get install and config steps with downloadable scripts, then add channels or skills from the same flow.

Before you begin, ensure you have:

• Node.js 22+ installed (or use the one-liner installer which includes it)
• An AI model subscription (Anthropic Claude Pro/Max recommended, or OpenAI ChatGPT Plus)
• macOS, Linux, or Windows (Windows users should use WSL2)
• Terminal/Command Prompt access

We'll use the one-liner installation method, which is the easiest for beginners. It automatically installs Node.js and all dependencies.

curl -fsSL https://openclaw.ai/install.sh | bash

What this does: This single command downloads and runs the OpenClaw installer, which will:

• Check if Node.js is installed (installs it if missing)
• Install OpenClaw globally via npm
• Set up the necessary dependencies

Platform support: Works on macOS, Linux, and Windows (via Git Bash or WSL2).

Verify Installation

After installation completes, verify it worked:

openclaw --version

You should see a version number (e.g., 2026.1.29). If you get "command not found", try restarting your terminal or check that Node.js is in your PATH.

The onboarding wizard is the easiest way to set up OpenClaw. It provides an interactive Terminal User Interface (TUI) that guides you through the entire setup process.

openclaw onboard --install-daemon

The wizard will launch an interactive terminal UI and guide you through:

• Confirmation to continue setup
• Onboarding mode selection - Choose QuickStart (recommended for beginners) which configures everything with safe defaults
• Gateway configuration and local setup
• Model selection and authentication
• Channel connections (Telegram, WhatsApp, Discord, etc.)
• Skill installation (optional)
• API key configuration (optional)

After QuickStart completes, the wizard will prompt you to set up your AI model. The wizard will show you a list of available model providers.

Supported Model Providers

OpenClaw supports multiple AI model providers:

• Anthropic Claude - Recommended for best quality (Pro/Max subscription with OAuth)
• OpenAI - ChatGPT Plus or API access
• Google Gemini - Via Google Antigravity OAuth (requires Gmail account)
• Other providers - Gemini, Moonshot, Minimax, and more

Authentication Methods

Depending on your chosen provider, you'll authenticate using:

• OAuth - For subscription services (Claude Pro/Max, ChatGPT Plus, Google Gemini). The wizard will open a browser for authentication.
• API Keys - For pay-per-use services. You'll paste your API key when prompted.

Recommended: Anthropic Claude

We recommend Anthropic Claude Pro/Max with Opus 4.6 because it offers:

• Excellent long-context handling for complex conversations
• Better protection against prompt injection attacks
• High-quality, nuanced responses
• OAuth authentication (no API keys needed with subscription)

After configuration: Your settings are saved to ~/.clawdbot/moltbot.json (or ~/.openclaw/ for newer installations). For advanced configuration options, see the Configuration Reference.

After setting up your model and channels, the wizard will ask you to choose how you want to interact with OpenClaw:

For this tutorial, select "Hatch in TUI" to continue with the terminal interface.

When you select "Hatch in TUI", the wizard will ask you some simple questions to personalize your agent:

• Bot's name - What should your assistant be called? (e.g., "Becky", "Alex", "Assistant")
• What it can call you - How should the bot address you? (e.g., your first name, "User", "Boss")

After answering these questions, your OpenClaw will confirm its setup with a personalized response that reflects the personality you've configured. This is your agent's "hatching" moment!

If you used --install-daemon, the Gateway should already be running. Verify it's active:

openclaw gateway status

If the Gateway isn't running, start it manually:

openclaw gateway --port 18789 --verbose

You should see output indicating the Gateway is running and listening on port 18789.

Now let's verify everything works by sending a test message:

Option A: Via CLI

openclaw agent --message "Hello! Can you introduce yourself?"

Option B: Via Connected Channel

If you set up a channel (Telegram, WhatsApp, etc.) during onboarding, simply send a message to your bot. It should respond!

Now that you have OpenClaw running, here's what to explore next:

If you encounter issues:

• Gateway won't start: Check if port 18789 is already in use: lsof -i :18789
• Authentication errors: Verify your API keys or OAuth tokens in the configuration file
• Model not responding: Check your model subscription status and API limits
• Installation issues: Ensure Node.js 22+ is installed: node --version

For more help, see our Troubleshooting Guide or FAQ.