Installing OpenClaw: A Complete Guide from Scratch

You've probably used ChatGPT, Claude, or Gemini. They're convenient, but there's a catch: your conversations live on someone else's servers. Every prompt, every personal detail, every half-baked idea goes through their infrastructure.

OpenClaw takes a different approach. It's an open-source AI assistant that runs on your own hardware. You pick the model, set the rules, and decide where your data goes.

Why Run Your Own AI Assistant?

A few reasons people do this:

Privacy (with local models): Run Ollama or similar, and your conversations never leave your network. No training on your data, no retention policies.
Cost control: Use local models like Qwen3 or Llama for free, or mix cloud APIs when you need more power.
Integration: OpenClaw connects to Telegram, WhatsApp, Discord, Slack, Signal. It becomes your personal assistant across all platforms.
Customization: Define your assistant's personality, give it access to your files, let it run shell commands, schedule tasks.
No limits: No "you've reached your daily limit" messages. No subscription tiers. It's your hardware.

One important caveat: if you use cloud APIs like Anthropic, OpenAI, or Google Gemini, your prompts still go through their servers. For true privacy, you need local models. The good news? Models like Qwen3 and Llama run great on modest hardware, and OpenClaw makes it easy to switch between local and cloud depending on the task.

Think of it as having your own Jarvis, running 24/7 on a Raspberry Pi, a homelab VM, or a cloud VPS. It remembers context across sessions, can browse the web for you, manage your calendar, send messages on your behalf.

What We're Building

This guide covers a production-ready setup: OpenClaw on Ubuntu, HAProxy for HTTPS, and Ollama for local AI models. When you're done:

What This Setup Gives You

Once done:

✅ A fully functional AI assistant accessible via HTTPS
✅ Web-based Control UI for managing your assistant
✅ Telegram bot integration
✅ Secure authentication with token-based access
✅ Professional reverse proxy setup with HAProxy
✅ Local LLM support with Ollama and Qwen3:32b

Prerequisites

System Requirements

Ubuntu Server 22.04 or 24.04
4 vCPUs minimum
8GB RAM (32GB+ if running local models)
20GB disk (100GB+ for local models)

External Requirements

Anthropic API key from console.anthropic.com (or use Ollama for local models)
HAProxy set up (optional, for HTTPS)
A domain pointing to your server (optional)

Part 1: System Preparation

Step 1: Update Your System

sudo apt-get update
sudo apt-get upgrade -y

Step 2: Install Node.js 22+

OpenClaw needs Node.js 22+:

# Add NodeSource repository
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -

# Install Node.js
sudo apt-get install -y nodejs

# Verify installation
node --version  # Should show v22.x.x
npm --version

Step 3: Configure npm for User-Level Global Installs

# Create directory for global packages
mkdir -p ~/.npm-global

# Configure npm to use it
npm config set prefix '~/.npm-global'

# Add to PATH
echo 'export PATH=~/.npm-global/bin:$PATH' >> ~/.bashrc
source ~/.bashrc

Part 2: Installing OpenClaw

Step 4: Install OpenClaw

npm install -g openclaw@latest

Step 5: Verify Installation

openclaw --version

Terminal showing OpenClaw installation progress

Part 3: Configuration Wizard

Step 6: Generate Your Authentication Token

openssl rand -hex 32

Save this token securely - you'll need it to access the web interface!

Step 7: Run the Onboarding Wizard

openclaw onboard --install-daemon

Follow the prompts:

Gateway Location: Select Local
Gateway Auth: Select Token (recommended)
Model Provider: Select Anthropic (or Ollama for local)
Primary Model: claude-sonnet-4-5 or local model

Part 4: Configuring Ollama for Local LLM Access

Want to run models locally instead of paying for API calls? Ollama makes it easy.

Step 8: Install Ollama

curl -fsSL https://ollama.com/install.sh | sh

Step 9: Configure Ollama for Network Access

Ollama only listens on localhost by default. To let OpenClaw connect from elsewhere:

# Edit the systemd service
sudo systemctl edit ollama.service

Add this content:

[Service]
Environment="OLLAMA_HOST=0.0.0.0"

Restart Ollama:

sudo systemctl daemon-reload
sudo systemctl restart ollama

Step 10: Pull Qwen3:32b Model

ollama pull qwen3:32b

That downloads about 20GB. If you want something lighter:

ollama pull qwen3:8b    # Faster, less resource-intensive

Step 11: Configure OpenClaw to Use Ollama

Edit your OpenClaw configuration:

nano ~/.openclaw/openclaw.json

Add Ollama as a provider in the auth.profiles section:

"auth": {
  "profiles": {
    "ollama:local": {
      "provider": "ollama",
      "mode": "none",
      "baseUrl": "http://YOUR-OLLAMA-IP:11434"
    },
    "anthropic:default": {
      "provider": "anthropic",
      "mode": "api_key"
    }
  }
}

Set your default model to use Ollama:

"agent": {
  "model": "ollama/qwen3:32b"
}

Step 12: Test Ollama Connection

# Verify Ollama is running
curl http://localhost:11434/api/tags

# Restart OpenClaw to apply changes
systemctl --user restart openclaw-gateway

Part 5: Configuration for Reverse Proxy

Step 13: Edit Configuration for HAProxy

nano ~/.openclaw/openclaw.json

Set up the gateway:

"gateway": {
  "port": 18789,
  "mode": "local",
  "bind": "lan",
  "controlUi": {
    "enabled": true,
    "allowInsecureAuth": true
  },
  "auth": {
    "mode": "token",
    "token": "YOUR-TOKEN-HERE"
  },
  "trustedProxies": [
    "HAPROXY-IP-1",
    "HAPROXY-IP-2"
  ]
}

Key settings:

bind: "lan" - Allows connections from your network
allowInsecureAuth: true - Required for proxy access
trustedProxies - Your HAProxy server IPs

Step 14: Restart the Gateway

systemctl --user restart openclaw-gateway
systemctl --user status openclaw-gateway

Part 6: HAProxy Configuration

Step 15: Configure HAProxy Backend

backend openclaw_backend
    description OpenClaw AI Assistant Gateway
    balance roundrobin
    option http-server-close
    option forwardfor
    timeout server 3600s
    timeout tunnel 3600s
    option httpchk GET /
    server openclaw01 192.168.0.243:18789 check inter 20s rise 2 fall 3 maxconn 50

Step 16: Configure HAProxy Frontend

frontend https_frontend
    bind *:443 ssl crt /etc/haproxy/certs/
    http-request set-header X-Forwarded-Proto https
    http-request set-header X-Forwarded-Host  %[req.hdr(Host)]
    http-request set-header X-Forwarded-For   %[src]
    
    acl host_openclaw hdr(host) -i your-domain.com
    use_backend openclaw_backend if host_openclaw

Step 17: Reload HAProxy

sudo systemctl reload haproxy

Part 7: Testing Your Installation

Step 18: Test Access

# Test local access
curl -I http://192.168.0.243:18789

# Test Ollama model
curl http://localhost:11434/api/generate -d '{
  "model": "qwen3:32b",
  "prompt": "Hello!"
}'

Step 19: Access the Web UI

Open your browser:

https://your-domain.com/?token=YOUR-TOKEN-HERE

Part 8: Telegram Integration (Optional)

Step 20: Configure Telegram Bot

Open Telegram and search for your bot
Send /start to the bot
Approve pairing:

openclaw pairing list
openclaw pairing approve telegram <CODE>

Security Tips

Strong Tokens: Use openssl rand -hex 32
HTTPS Only: Never expose over plain HTTP publicly
Network Restrictions: Use HAProxy ACLs to restrict access
Keep Updated: npm update -g openclaw
Backup Config: cp ~/.openclaw/openclaw.json ~/.openclaw/openclaw.json.backup

Resources

Official Documentation: https://docs.openclaw.ai
GitHub Repository: https://github.com/openclaw/openclaw
Discord Community: https://discord.gg/clawd
Ollama Documentation: https://ollama.com/library

That's the whole setup. You now have OpenClaw running with both cloud and local options. Use Anthropic or OpenAI when you need top-tier quality. Switch to Ollama when privacy matters or you just don't want to pay per token.

Things to remember:

OpenClaw requires Node.js 22+
allowInsecureAuth: true is required when running behind a proxy
Ollama works great for running local models like Qwen3
Configure OLLAMA_HOST=0.0.0.0 for network access