OpenClaw on a VPS

Running OpenClaw on a VPS gives you 24/7 availability without keeping a home machine on. The good news: OpenClaw is lightweight on the server side. The AI computation happens on your LLM provider (OpenAI, Anthropic, Ollama cloud, etc.), so your VPS mainly runs Node.js, channel connections, and the gateway. That means a modest VPS—1–2GB RAM, single vCPU—is usually enough for personal or small-team use.

This guide covers minimum requirements, recommended providers, and step-by-step setup. Compare Hetzner (EU value), DigitalOcean, Vultr, Oracle free tier, and managed options below.

Pick a datacenter close to you for SSH and WebChat—and think about where your LLM API runs (Anthropic, OpenAI, Bedrock). The Gateway is light; latency you feel is usually model API round-trips, not the VPS CPU. Workplace bots on Slack or Teams still need a reachable Gateway: Slack quickstart, Teams quickstart. Hands-off hosting: managed providers.

There is no single "best" host—match budget, region, and whether you need HTTPS for Teams. This table is a decision aid; step-by-step guides are on each provider page.

Slack bots often use Socket Mode (no public inbound port). Microsoft Teams needs a public HTTPS messaging URL—plan TLS on the VPS or use a tunnel (remote access). See Teams quickstart.

OpenClaw needs:

• RAM: 1GB minimum; 2–4GB recommended for comfortable operation (multiple channels, skills, browser automation)
• CPU: 1 vCPU is enough for typical chat and automation; 2 vCPU helps with concurrent sessions
• Storage: 10–20GB for base install, workspace, and logs; more if you run Ollama locally
• Node.js: 22.19 or later (24 recommended)

Important: If you use cloud LLMs (Anthropic, OpenAI, etc.), the VPS doesn't run the models. Heavy lifting happens on their servers. Your VPS only handles the Gateway, channel connections, and tool execution. That's why a small VPS works.

Here are the top providers for OpenClaw, with real-world pricing and notes:

Once you have a VPS with Ubuntu or Debian:

1. Install Node.js 20+ (or use the OpenClaw one-liner, which installs Node if needed)
2. Install OpenClaw and PM2
3. Run the onboarding wizard
4. Start the gateway with PM2 so it survives reboots

curl -fsSL https://openclaw.ai/install.sh | bash
npm install -g pm2
openclaw init
pm2 start 'openclaw gateway start' --name openclaw
pm2 save
pm2 startup

Enable UFW firewall for ports 22 (SSH), 80 (HTTP), and 443 (HTTPS). For full setup including reverse proxy and SSL, see our Deployment tutorial and the provider-specific guides below.

For reproducible deployments, use Docker. The Docker guide covers building and running OpenClaw in containers. Many providers (DigitalOcean, Hetzner) support Docker out of the box. Benefits: consistent environment, easy updates, isolation.

Running OpenClaw on a local device (your Mac, Linux box, or Windows machine) often improves day-to-day productivity: you can airdrop files, use local paths, and get lower latency. Workflows like "drop a video, get transcript and thumbnails" are much easier when the Gateway and workspace live on a machine you use daily.

Use a VPS when you need 24/7 reachability (e.g. WhatsApp always online), don't have suitable local hardware, or want to isolate the assistant from your main machine. Prefer no servers at all? See Managed hosting—Clawi, xCloud, MyClaw.ai let you get WhatsApp and Telegram connected without running any infrastructure. More in Tips & best practices.