Running OpenClaw on a VPS

VPS checklist

1. Pick a size — The agent itself is light; browser automation and code execution need more memory. 2 GB RAM is a practical minimum if you use the browser.
2. Create a non-root user — Run the agent under its own user account with SSH keys only.
3. Install Node with a version manager — Avoid global sudo installs; match the Node version the OpenClaw docs require.
4. Install and onboard OpenClaw — Install the CLI, run onboarding and install the daemon so it survives reboots.

   npm install -g openclaw@latest
   openclaw onboard --install-daemon

5. Lock it down — Enable a firewall, keep the OS updated, and sandbox command execution in Docker.
6. Monitor — Set up uptime alerts and log rotation so you know when the gateway stops.

Prefer not to run a server?

MobiClaw runs OpenClaw-style agents for you with sandboxing, browser automation, scheduling and chat-app connections already configured. You get the always-on agent without the sysadmin work.

Frequently asked questions

Do I need a GPU to host an AI agent?

No. The agent calls hosted AI models over the internet, so a regular CPU server is enough.

Is there a free trial?

Yes. Every paid plan starts with a 24-hour trial for $1 so we can keep abuse out. If you do nothing, the subscription starts automatically after the trial; cancel any time before that from Billing and you will not be charged again.