Home › Getting Started

Getting Started with OpenClaw

Your free guide to installing, configuring, and running your first personal AI agent. From zero to a working agent in under an hour.

What is OpenClaw?

OpenClaw is an open-source platform for building personal AI agents. Unlike cloud-based chatbots like ChatGPT or Claude.ai, an OpenClaw agent runs on your hardware — a Mac Mini, a Linux server, or even a $5/month VPS. It connects to your messaging apps (WhatsApp, Telegram, Discord, Slack, Signal, iMessage) and acts as your personal AI assistant across every platform you use.

What makes OpenClaw different is control. You define your agent's personality through a file called SOUL.md. You manage its memory. You choose which AI models it uses (GPT-5, Claude, Gemini, DeepSeek — or run local models). You decide what tools it can access and what it can do autonomously. It's the difference between renting an apartment and owning a house.

Step 1: Install OpenClaw

OpenClaw supports multiple installation methods. Choose the one that fits your setup:

Homebrew (macOS / Linux)

# Install OpenClaw via Homebrew brew install openclaw # Verify installation openclaw --version

npm (Any platform with Node.js)

# Install globally via npm npm install -g openclaw # Or use npx without installing npx openclaw --version

Docker

# Pull and run the official Docker image docker pull openclaw/gateway docker run -d --name openclaw -p 3000:3000 openclaw/gateway

For detailed installation instructions including troubleshooting, see our complete installation tutorial.

Step 2: Start the Gateway

The gateway is OpenClaw's core daemon — it manages your agent, routes messages between channels, and handles AI model communication.

# Start the gateway openclaw gateway start # Check status openclaw gateway status # View logs openclaw gateway logs

Step 3: Configure Your AI Model

OpenClaw needs at least one AI model provider. The most common setups:

# Using OpenAI openclaw config set model.provider openai openclaw config set model.apiKey sk-your-key-here openclaw config set model.default gpt-4 # Using Anthropic (Claude) openclaw config set model.provider anthropic openclaw config set model.apiKey sk-ant-your-key-here openclaw config set model.default claude-sonnet-4-6 # Using multiple providers (route by task) openclaw config set model.routing.default anthropic/claude-opus-4-6 openclaw config set model.routing.code openai/gpt-5 openclaw config set model.routing.simple anthropic/claude-haiku-4

For a full breakdown of model providers and routing strategies, see the Complete Guide or Chapter 9 of the book.

Step 4: Create Your First SOUL.md

SOUL.md is the most important file in your agent's workspace. It defines personality, communication style, values, and decision-making behavior. Here's a starter template:

~/.openclaw/workspace/SOUL.md
# SOUL.md — Your Agent's Personality ## Identity You are a personal AI assistant. Your name is [Agent Name]. You were configured by [Your Name] to help with daily tasks, communication, and information management. ## Personality - Tone: Professional but approachable - Communication: Clear, concise, action-oriented - Humor: Light when appropriate, never forced - Proactivity: Suggest improvements, don't wait to be asked ## Values - Accuracy over speed - Privacy first — never share personal data - Complete tasks, don't just acknowledge them - When uncertain, ask rather than assume ## Communication Style - Use short paragraphs - Lead with the answer, then explain - Use bullet points for multiple items - Code blocks for technical content

This is just the beginning. See our complete SOUL.md guide for advanced personality crafting, or check the Resources page for three different SOUL.md templates.

Step 5: Create Your First MEMORY.md

MEMORY.md is your agent's long-term memory — curated information that persists across sessions. Unlike daily memory files (which are automatic logs), MEMORY.md contains intentionally saved context.

~/.openclaw/workspace/MEMORY.md
# MEMORY.md — Long-Term Memory ## About My Human - Name: [Your Name] - Role: [Your job/role] - Timezone: [Your timezone] - Communication preference: [e.g., "concise, no fluff"] ## Key Decisions [Agent will populate this as you work together] ## Learned Preferences [Agent will learn these over time] ## Active Projects [Agent will track these]

Your agent will grow this file over time. The memory system is covered in depth in our Memory System Explained article.

Step 6: Connect Your First Channel

The magic of OpenClaw is that your agent lives in your messaging apps. Start with one channel:

# Connect Telegram (easiest to start with) openclaw channel add telegram # Follow the prompts to enter your bot token from @BotFather # Connect WhatsApp openclaw channel add whatsapp # Scan the QR code with your phone # Connect Discord openclaw channel add discord # Enter your bot token from Discord Developer Portal

For detailed channel setup guides, see:

Next Steps

You now have a working personal AI agent. Here's where to go from here:

📖 Go Deeper with the Book

This getting started guide covers the basics. The Personal Agent Revolution covers everything — 37 chapters, 187 pages, from beginner to advanced. Plus 3 bonus resources including the Agent Prompt Vault (50+ templates) and the Automation Playbook (30 ready-made automations).

Get the Book — $29.95 →

30-day money-back guarantee · Instant delivery · Lifetime updates

FAQ

Yes. OpenClaw is MIT-licensed open source software. The gateway is completely free. You still need to pay for AI model API access (OpenAI, Anthropic, etc.) or use a subscription like Claude Max or ChatGPT Plus.
macOS, Linux, and Windows (via WSL). The most common setup is a Mac Mini or Linux server running 24/7, but you can also use a cloud VPS.
No. OpenClaw connects to cloud AI APIs (OpenAI, Anthropic, etc.) — the heavy computation happens on their servers. Your gateway is just a lightweight Node.js application. If you want to run local models with Ollama, a GPU helps but isn't required for smaller models.