OpenClaw Setup Guide: From Install to First Agent

⚠️ Before You Start OpenClaw requires technical setup including Node.js, environment configuration, and API keys. Most businesses benefit from professional setup to ensure security and optimal configuration. This guide covers the basics — for production deployment, consider our setup service.

Prerequisites

Before installing OpenClaw, ensure you have:

Node.js 18+ — Download from nodejs.org
npm or yarn — Comes with Node.js
Git — For version control
API Keys — At minimum, an Anthropic or OpenAI key
Basic terminal knowledge — You'll be running commands

💡 Hardware Requirements OpenClaw runs on any modern machine. Recommended: 4GB RAM, 2GB storage for basic setup. For production with multiple agents and skills, 8GB+ RAM is better.

Installation

Install OpenClaw CLI

Install the OpenClaw command-line interface globally:

npm install -g openclaw

Verify installation:

openclaw version

Initialize Your Workspace

Create a new OpenClaw workspace:

mkdir my-openclaw-setup
cd my-openclaw-setup
openclaw init

This creates the basic directory structure and configuration files.

Configure Environment Variables

OpenClaw needs API keys and configuration. Create a .env file:

# AI Model Keys
ANTHROPIC_API_KEY=your_anthropic_key_here
OPENAI_API_KEY=your_openai_key_here

# Gateway Configuration  
OPENCLAW_PORT=3000
OPENCLAW_HOST=localhost

# Optional: Telemetry (set to false to disable)
OPENCLAW_TELEMETRY=false

🔐 Security Note Never commit API keys to version control. Add .env to your .gitignore file immediately.

Create Your First Agent

Generate an agent configuration:

openclaw agent create

You'll be prompted to configure:

Agent name (e.g., "Assistant", "Ops Agent")
Primary model (Claude, GPT-4, etc.)
System prompt/persona
Initial skills to include

Start the Gateway

Launch the OpenClaw gateway server:

openclaw gateway start

You should see output like:

✓ Gateway running on port 3000
✓ Agent "Assistant" connected  
✓ Skills loaded: 3
✓ Ready to receive connections

Test Your Agent

Open a new terminal and test the connection:

openclaw chat

This opens a direct chat interface. Try asking:

"Hello, who are you?"
"What skills do you have?"
"What can you help me with?"

✅ Success! If your agent responds, congratulations — OpenClaw is working. Your AI agent is now running locally and ready for configuration.

Basic Configuration

Agent Persona (SOUL.md)

Your agent's personality is defined in the SOUL.md file. This controls how it communicates and behaves:

# SOUL.md — Your Assistant

**Name:** Assistant | **Email:** assistant@yourcompany.com 
**Emoji:** 🤖 | **Persona:** Professional helper

## Tone & Style  
- Professional but friendly
- Clear, concise responses
- Always ask for clarification when uncertain
- Use bullet points for complex information

## Capabilities
- Answer questions about your business
- Help with scheduling and organization  
- Provide information and research
- Create content and documents

## Never
- Make assumptions about sensitive topics
- Access external systems without permission
- Share confidential information

Memory Configuration

OpenClaw agents maintain memory in two forms:

Daily Memory: memory/YYYY-MM-DD.md — Detailed daily logs
Long-term Memory: MEMORY.md — Curated insights

The agent automatically manages these files, but you can review and edit them to guide future behavior.

Adding Skills

Skills extend your agent's capabilities. Install your first skill:

openclaw skills install web-search

Available built-in skills include:

web-search — Search the internet
file-manager — Read/write local files
calendar — Schedule management
email — Email integration
github — Git repository management

Connecting Clients

Mobile App

Download the OpenClaw companion app:

iOS: Search "OpenClaw" in the App Store
Android: Download from Google Play

Connect by scanning the QR code displayed when you run:

openclaw node pair

Telegram Integration

For Telegram integration:

Create a bot via @BotFather
Get the bot token
Add to your .env:

TELEGRAM_BOT_TOKEN=your_bot_token_here

Restart the gateway and your agent will be available in Telegram.

Common Issues & Troubleshooting

Gateway Won't Start

Port in use: Change OPENCLAW_PORT in .env
Missing API keys: Verify all required keys are in .env
Node version: Ensure Node.js 18+ is installed

Agent Not Responding

Model issues: Check API key validity and account credits
Rate limiting: Reduce request frequency
Memory issues: Clear the memory/ directory if corrupted

Skills Not Loading

Permission errors: Ensure proper file permissions
Dependency issues: Run npm install in the skills directory
Configuration errors: Check skill YAML files for syntax

Next Steps

Now that OpenClaw is running, consider:

Custom Skills: Build skills specific to your business workflows
Automation: Set up cron jobs for recurring tasks
Security: Implement proper access controls for production
Monitoring: Add health checks and logging
Integration: Connect to your existing tools and APIs

Production Deployment

For production use, you'll need:

Dedicated server — VPS, cloud instance, or on-premise
Process management — PM2, systemd, or similar
Reverse proxy — nginx for HTTPS and domain routing
Backup strategy — Regular memory and configuration backups
Monitoring — Health checks, logging, and alerting
Security hardening — Firewall, access controls, key rotation

⚠️ Production Considerations Setting up OpenClaw for production involves security, reliability, and performance considerations beyond this basic guide. We recommend professional setup for business-critical deployments.

Getting Help

If you're stuck:

Documentation: Check the official OpenClaw docs
Community: Join the OpenClaw Discord/forums
Professional Setup: Book a consultation with our team

Remember: OpenClaw is powerful but complex. Don't hesitate to get help — a properly configured system saves weeks of troubleshooting later.

OpenClaw Setup