How to Install OpenClaw on a VPS: Complete Setup Guide (2026)

Running OpenClaw on a VPS means your AI assistant is always available — whether your laptop is closed, on a plane, or just turned off. Unlike a local Mac install, a VPS setup gives you 24/7 uptime, a static IP, and the ability to receive webhooks and messages at any hour.

This guide walks you through a full OpenClaw VPS installation from scratch, using a clean Ubuntu 22.04 server. The same steps work on Hetzner, DigitalOcean, Vultr, Linode, or any comparable provider.

Why run OpenClaw on a VPS?

A local install on your Mac is great for getting started, but a VPS unlocks the full potential of OpenClaw:

• Always on — receives Telegram/WhatsApp messages even when your machine is off
• Stable endpoint — webhooks, n8n integrations, and bots all need a reliable URL
• No performance hit — offloads AI calls and automation workloads from your laptop
• Team-ready — easier to share access and add team members later

A $6/month Hetzner CAX11 (2 vCPU, 4GB RAM) is more than enough to run OpenClaw, its gateway, and a few connected automations.

What you need before starting

• A VPS running Ubuntu 22.04 LTS (or 20.04) — root or sudo access
• An AI API key — Claude (Anthropic), GPT-4o (OpenAI), or Gemini
• A Telegram bot token (optional but recommended — create one via BotFather)
• ~15 minutes of uninterrupted time

Step 1 — Provision your server

Create a new VPS with at least 2GB RAM and 20GB disk. Ubuntu 22.04 LTS is the recommended OS. Once it's up, SSH in:

ssh root@YOUR_SERVER_IP

Update the system first:

apt update && apt upgrade -y

Step 2 — Install Node.js

OpenClaw requires Node.js 22 LTS or higher. Install it using the NodeSource setup script:

curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
apt install -y nodejs

Verify it installed correctly:

node --version   # should print v22.x.x
npm --version

Step 3 — Install OpenClaw

Install OpenClaw globally via npm:

npm install -g openclaw

This installs the openclaw CLI. Verify it works:

openclaw --version

Step 4 — Run initial setup

OpenClaw has an interactive onboarding wizard. Run it once to set your AI provider, API key, and workspace:

openclaw setup

The wizard will ask you to:

1. Choose your AI model provider (Claude, GPT-4o, Gemini, Ollama)
2. Paste your API key
3. Set a workspace directory (default: ~/.openclaw/workspace)
4. Optionally configure a persona name and initial memory

Your config is stored at ~/.openclaw/config.json. You can edit it directly later.

Step 5 — Configure the gateway

The OpenClaw gateway is what allows messaging channels (Telegram, WhatsApp, etc.) to reach your assistant. Start it:

openclaw gateway start

Check that it's running:

openclaw gateway status

By default, the gateway listens on port 3210. If you have a firewall, open that port:

ufw allow 3210/tcp

Step 6 — Keep it always-on with systemd

To make OpenClaw survive reboots and run as a background service, create a systemd unit file:

cat > /etc/systemd/system/openclaw.service << 'EOF'
[Unit]
Description=OpenClaw AI Assistant
After=network.target

[Service]
Type=simple
User=root
ExecStart=/usr/bin/openclaw gateway start --foreground
Restart=on-failure
RestartSec=10
StandardOutput=journal
StandardError=journal

[Install]
WantedBy=multi-user.target
EOF

Enable and start the service:

systemctl daemon-reload
systemctl enable openclaw
systemctl start openclaw
systemctl status openclaw

Your OpenClaw gateway will now start automatically on boot and restart if it crashes.

Step 7 — Connect Telegram

This is the most satisfying part — chatting with your assistant from your phone. You'll need a Telegram bot token from BotFather.

Add the token to your OpenClaw config:

openclaw config set channels.telegram.token YOUR_BOT_TOKEN

Then restart the gateway:

systemctl restart openclaw

Open Telegram, find your bot, and send it a message. You should get a response from your AI assistant within a few seconds. For a full walkthrough, see our Telegram setup guide.

Troubleshooting common issues

Gateway won't start

Check the logs: journalctl -u openclaw -n 50. The most common causes are a missing API key, port conflict (something already using 3210), or a Node.js version below 22.

Telegram bot not responding

Make sure your gateway is reachable from the internet on port 3210. Test with: curl http://YOUR_SERVER_IP:3210/health. If it times out, check your VPS firewall rules.

High memory usage

OpenClaw itself is lightweight (~100MB), but repeated large AI conversations can grow memory over time. If you're on a 1GB RAM server, consider adding a 1GB swap file.

npm install fails

If you get permission errors, make sure you're running as root, or use sudo npm install -g openclaw. On non-root setups, configure npm's global prefix to a user-writable directory.

What to do next

Once your VPS setup is working, here's what most people do next:

• Add a domain — point a subdomain at your VPS IP for a cleaner webhook URL
• Connect email & calendar — let your assistant read and respond to your inbox
• Set up n8n — pair OpenClaw with n8n for powerful workflow automations
• Configure memory — teach your assistant about your projects, preferences, and context
• Add skills — extend OpenClaw with community skills from ClawHub

If you'd rather skip the terminal entirely and have us handle this for you, our Power User setup includes VPS deployment, all channel connections, and 60 days of support — starting at $899.