How to host OpenClaw 24/7

Short answer. OpenClaw runs as one long-lived process, the Gateway, so to keep your agent answering 24/7 it needs a machine that stays on all the time. OpenClaw's own docs recommend a VPS for that: Ubuntu LTS, Node.js 24.16+ or 26.1+, the installer script, then openclaw onboard --install-daemon, which installs the Gateway as a systemd service that restarts itself. Keep the Gateway on loopback and reach its dashboard through an SSH tunnel. Or rent a server that runs OpenClaw for you and talk to it from Telegram or Discord.

Checked on 6 November 2026, accurate for OpenClaw 2026.11.1.

This guide uses OpenClaw's own documentation (docs.openclaw.ai) and what we measured running OpenClaw 2026.11.1 on our own servers. To go straight to the rental route, see Option 2.

Why a laptop is not enough

OpenClaw answers your messages and runs its scheduled jobs only while the Gateway runs. OpenClaw's FAQ puts it plainly: for 24/7 reliability, use a VPS; a laptop is fine if you accept that sleep, network drops, OS updates and reboots disconnect it. Telegram, Discord, Slack and the other chat apps all work from a VPS.

What you need

Our 2 GB numbers were taken with no AI model answering and no chat app in use; real use needs more, which is why our panel recommends 4 GB. You also need an API key from an AI provider (see which AI provider key to use) and a chat app to talk to it, such as a Telegram or Discord bot. More on memory in how much memory OpenClaw needs.

Option 1: host it yourself on a VPS

These steps follow OpenClaw's DigitalOcean guide, which works the same on any Ubuntu 24.04 LTS server.

1. Create the server from a clean Ubuntu 24.04 LTS image, with an SSH key. OpenClaw's guide warns against third-party one-click images unless you have read their startup scripts and firewall defaults.
2. On a 1 GB server, add swap first. OpenClaw's guide recommends a 2 GB swap file for its 1 GB plan, and in our test the npm install alone peaked at about 1.6 GB of memory:

   sudo fallocate -l 2G /swapfile
   sudo chmod 600 /swapfile
   sudo mkswap /swapfile
   sudo swapon /swapfile
   echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab

3. Install Node.js and OpenClaw as root:

   apt update && apt upgrade -y
   curl -fsSL https://deb.nodesource.com/setup_24.x | bash -
   apt install -y nodejs
   curl -fsSL https://openclaw.ai/install.sh | bash -s -- --no-onboard

4. Create a normal user that owns OpenClaw, and let its services keep running after you log out:

   adduser openclaw
   usermod -aG sudo openclaw
   loginctl enable-linger openclaw
   su - openclaw
   openclaw --version

5. Run onboarding and install the service:

   openclaw onboard --install-daemon

   The wizard asks for your model provider, sets up channels, generates the Gateway token and installs a systemd user service. The unit it writes uses Restart=always, so the Gateway comes back after a crash, and the linger step above lets it start again after a reboot without you logging in.
6. Check that it runs:

   openclaw status
   systemctl --user status openclaw-gateway.service
   journalctl --user -u openclaw-gateway.service -f

Open the dashboard safely

The Gateway listens on port 18789 and binds to loopback by default, so it is not reachable from the internet. Do not open port 18789 in your firewall. From your own computer, open a tunnel and leave it running:

ssh -L 18789:localhost:18789 root@YOUR_SERVER_IP

Then open http://localhost:18789 and sign in with the Gateway token (on the server, openclaw gateway auth-token --show prints it). OpenClaw also documents Tailscale Serve as a second option.

Keep it safe and up to date

• Who can message it. New Telegram and Discord DMs start in pairing mode: a stranger gets a code and is ignored until you approve it with openclaw pairing approve telegram <CODE>. Pairing codes expire after one hour. Never switch a personal agent to open: anyone who finds the bot could then command it.
• One owner per Gateway. OpenClaw's security model is one trust boundary per Gateway. People who do not trust each other need separate Gateways, ideally on separate OS users or hosts.
• Updates. OpenClaw releases often. Run openclaw update: it checks the new version while the old Gateway keeps serving, then switches over and verifies it.
• Backups. Everything lives under ~/.openclaw/ (config, credentials, SQLite stores, the agent's workspace and memory). openclaw backup create makes an archive. Treat that folder and its backups as secrets: OpenClaw's docs say OAuth tokens are stored as plain text in its SQLite files.

Prefer containers? See how to run OpenClaw with Docker.

Option 2: host it on PlayNHost

Here the server already runs OpenClaw. There is no Node.js to install, no service to write and no port to open: you paste your keys in the panel and press Start.

1. Pick a plan on the OpenClaw hosting page (2 GB, 4 GB or 8 GB of memory) and press Create Server.
2. Wait for the install. The server appears at panel.playnhost.com. While it installs, the Console ends with: "[PlayNHost] OpenClaw is ready to start. Add an AI provider key and a Telegram or Discord bot token on the Agent page, then press Start."
3. Open the Agent tab. It has three cards:
   • AI provider: one key is enough, from OpenAI, Anthropic, Google Gemini, xAI or OpenRouter. Press Set, paste it and press Save.
   • Chat apps: a Telegram bot token, a Discord bot token, or both.
   • Who can message it: your Telegram user ID and/or your Discord user ID. Until you add yours, nobody can message the agent.
   Keys and tokens are stored encrypted, never shown again, and reach OpenClaw through its environment.
4. Press Start. The first start installs OpenClaw. In our test of the same start script, the console printed "Your agent is running" 67 seconds after Start, install included, followed by a line saying who can message your agent.
5. Message your bot from Telegram or Discord. Step-by-step: connect OpenClaw to Telegram, connect OpenClaw to Discord.

What the panel does for you

• Updates at every start. With OpenClaw version set to latest (Agent page, Code, Settings), each start checks npm and installs a newer release when there is one. Enter an exact release such as 2026.11.1 to stay on it.
• Restarts. The Agent page's Restarts pane says: "If your bot stops by itself, we start it again within about a minute." If it crashes again straight after that restart, it is left stopped instead of restarting over and over.
• No open port. The Gateway listens only inside your server, on port 18789 of its loopback address, with a token the panel generates. Your agent connects out to Telegram and Discord on its own.
• Only you. Direct messages are limited to the user IDs on the Agent page. While that list is empty, nobody can message the agent: on Telegram, a first message is answered with your user ID so you can add it. If the settings are changed to let anyone in while the agent runs, the agent is stopped, and the next start puts your IDs back.
• Files and backups. The agent's memory and settings live in .openclaw, which you can see in Files. In our test the backup set was about 4 MB: the program and caches are reinstalled at start, not backed up.

What it does not do yet

• No web dashboard. You use OpenClaw from Telegram or Discord. OpenClaw's browser dashboard is not offered yet.
• Telegram and Discord only. WhatsApp is not allowed on PlayNHost; if it is switched on, the start stops with a sentence that says so and the plugin is removed. Other chat apps are switched off.
• No browser automation. The server has no Chrome or Chromium, so tasks that need the agent to drive a real browser do not work.
• API keys only. The panel takes provider API keys. Sign-ins such as a ChatGPT or Claude subscription login are not offered.

VPS or PlayNHost: side by side

If you need WhatsApp, the dashboard or a browser, run OpenClaw on a VPS. If you want an agent you message from Telegram or Discord without looking after a server, create an OpenClaw server.

PlayNHost is not affiliated with the OpenClaw Foundation. OpenClaw is open-source software under the MIT licence.