OpenClaw Onboarding Wizard: Everything It Does Explained

The openclaw onboard command is the recommended starting point for every new OpenClaw installation. Rather than manually editing configuration files or guessing which settings matter, the onboarding wizard walks you through every critical decision in a structured, interactive format. This guide explains exactly what the wizard does at each step and what choices lead to the best outcomes.

Running the Wizard

After installing OpenClaw, activate your virtual environment and run:

source venv/bin/activate
python -m openclaw onboard

The wizard runs in your terminal and takes about 10–15 minutes to complete from scratch.

Stage 1: LLM Provider Configuration

The wizard first asks which LLM provider you want to use. For each provider, it:

1. Validates that your API key is correctly formatted
2. Makes a test API call to confirm connectivity
3. Displays the cost-per-1K-tokens for the selected model
4. Sets up the rate limiting configuration appropriate for your tier

Recommended choices: Anthropic Claude 3.5 Sonnet for best quality, DeepSeek V3 for best cost-to-performance, or OpenRouter if you want to switch between models.

Stage 2: Channel Setup

The wizard presents all available channels and walks through credential setup for each one you select:

• Telegram: Enter your bot token from @BotFather; wizard tests connectivity
• Discord: Provides OAuth URL; you authorize in browser; wizard stores token
• Slack: Similar OAuth flow; select which workspace
• WhatsApp: Guides you through WhatsApp Business API credential entry

You can skip channels and add them later through the configuration file or the openclaw channels add command.

Stage 3: Skill Installation

The wizard pulls a curated list of popular skills from ClawHub and lets you select any to install immediately. Recommended for most users:

• web-search: Internet search capability (uses Brave Search API)
• memory: Enhanced persistent memory with vector search
• calendar: Google Calendar integration

Skills are installed into ~/.openclaw/skills/ and activated immediately.

Stage 4: Security Configuration

The wizard walks through security hardening:

• Sets a random secret key for session management
• Configures the tools.allow file to restrict which system operations your agent can perform
• Enables the canvas host security fix (patching the pre-CVE-2026-25253 default configuration)
• Optionally sets up two-factor authentication for the web interface

Stage 5: Test Conversation

The wizard conducts a test conversation with your agent — sending a simple "hello" message and confirming a response arrives on your connected channels. This verifies the complete pipeline from input to LLM to output before you finish setup.

Post-Wizard Configuration

Everything the wizard configures is stored in ~/.openclaw/openclaw.json. You can manually edit this file to customize settings beyond what the wizard offers.

Frequently Asked Questions

Can I re-run the onboarding wizard?

Yes. Run python -m openclaw onboard --reset to reset your configuration and re-run from the beginning, or python -m openclaw onboard --llm to redo only the LLM section.

What if the wizard fails on the API key validation step?

Double-check for leading/trailing spaces in your key and confirm the key has not been revoked at your provider's console. Many failures are simply copy-paste whitespace issues.

Does the wizard install everything needed to run OpenClaw?

Yes, plus some optional components. After completing the wizard, python -m openclaw start is sufficient to run your agent.