Run OpenClaw without a server
Updated
Short answer: let someone else run the Gateway. Hosted OpenClaw on ElfHosted runs your OpenClaw Gateway 24/7 in its own isolated container, with ElfHosted single sign-on in front of the web UI, updates handled, and an AI model already set up, so the agent answers on its first run without an API key. It's $9 a month after a $1, 7-day trial.
Why does OpenClaw need a server at all?
OpenClaw is a personal AI assistant. The part that does the work is the Gateway, a single Node.js process that receives messages from your chat channels (WhatsApp, Telegram, Slack, Discord, Signal, iMessage and many more), routes them to the agent, runs its tools (browser, shell, sessions, cron) and sends the replies.
The Gateway has to be running for any of that to happen. If it lives on a laptop, your agent goes quiet when the lid closes, scheduled jobs don't fire, and webhooks from your channels have nowhere to land.
What are the options?
There are three ways to keep a Gateway running, and they trade cost against effort.
| Your own computer | A VPS you manage | A managed host | |
|---|---|---|---|
| Always on | Only while it's awake | Yes | Yes |
| Who installs and updates it | You | You | The host |
| Where the agent's shell and tools act | Your everyday machine | Your VPS | An isolated container |
| HTTPS for channels that need a public address | You arrange it | You arrange it | Included |
| Sandboxing and access control | You configure it | You configure it | Set up for you |
| AI model | Bring your own key | Bring your own key | Varies; included on ElfHosted |
| Cost | Nothing extra | The VPS, plus your time | A monthly subscription |
Your own computer is OpenClaw's default, and it's a fine way to try it. Remember the agent's shell runs with your permissions on the machine holding your files.
A VPS you manage fixes the uptime problem and moves the agent off your laptop. You're now the sysadmin: OS patches, firewall, TLS certificates, Docker if you want OpenClaw's sandboxing (it's off by default), and upgrades that sometimes need config fixed afterwards.
A managed host runs all of that for you. That's what ElfHosted does.
What do you get with hosted OpenClaw on ElfHosted?
- Always on. The Gateway runs around the clock, so scheduled tasks, webhooks and long-running workflows fire when they should.
- Its own container. The agent's shell and tools act inside that container, not on your devices or home network.
- Single sign-on in front. The web UI sits behind your ElfHosted login, the Control UI only accepts your own subdomain, and the Gateway has its own per-instance token.
- Config that heals itself.
openclaw doctor --fixruns on every start, so a config left stale by an upgrade gets repaired. - AI included. An ElfHosted model is configured as a provider from the start, on an allowance that suits everyday use. Bring your own OpenAI, Anthropic, Gemini or other compatible provider whenever you like, and your choice is never overwritten.
- Web UI and CLI. Both open from your dashboard, so you can run
openclaw onboardwithout installing anything. - Your files, reachable. Everything OpenClaw keeps lives under
config/openclaw/, which you can browse and edit in FileBrowser.
OpenClaw or Hermes?
If you're looking for a personal AI agent in the cloud and haven't settled on one, ElfHosted hosts two:
- OpenClaw has the widest set of chat channels and a large skills library.
- Hermes, from Nous Research, leans into self-improving skills, searchable memory across sessions and a browser dashboard with chat built in. See Hosted Hermes Agent.
Both cost the same, and both include AI. Already on OpenClaw? Hermes imports your OpenClaw configuration on first launch and leaves your OpenClaw instance untouched, so trying it side by side is easy.
Getting started
- Start a $1, 7-day trial of hosted OpenClaw.
- Open OpenClaw CLI from your dashboard and run
openclaw onboard. You can skip the model step, because the included AI already works. - Open the OpenClaw web UI. The first time a browser connects it asks for the Gateway token, which
openclaw gateway auth-token --showprints in OpenClaw CLI. - Connect your chat channels and start talking to your agent.
The OpenClaw app page has the details on tokens, approving browsers and using your own provider.
Frequently asked questions
How do I run OpenClaw without my own server?
Use a managed host. Hosted OpenClaw on ElfHosted runs the Gateway 24/7 in its own container, with single sign-on, updates handled and AI included. It's $9 a month after a $1, 7-day trial.
Can I run OpenClaw on my own computer instead?
Yes, and it costs nothing extra. The agent stops answering when the computer sleeps or restarts, and its shell and tools act on the machine you use every day.
Do I need an API key?
No. The included ElfHosted model works from the first run. Add your own provider any time for a favourite model or heavier use, billed by that provider directly.
Is it safer than running OpenClaw on my laptop?
The Gateway has shell access and reads messages from your channels, which is text from people who aren't you. On ElfHosted it runs in its own container, so that access stays inside the container rather than on your laptop or home network.
Can I move to Hermes later?
Yes. Hermes imports your OpenClaw configuration automatically on first launch.
About AI agent tools
OpenClaw is an infrastructure tool for building and running AI agents. ElfHosted provides hosting infrastructure only. You're responsible for how you configure and use your agent, including following the terms of any third-party service it talks to. See our Acceptable Use Policy.