Skip to content
Odinex Cloud
DELog inSign up
OpenClaw
AI

OpenClaw

Your own open-source personal AI assistant.

OpenClaw, the open-source personal AI assistant, on Ubuntu 24.04 with HTTPS set up for you. Connect Anthropic, OpenAI, Google Gemini or OpenRouter with your own API key from a setup page in the browser.

Deploy OpenClawGetting started

Opens the Odinex console with OpenClaw selected. No licence fee: you pay for the server only.

Base image
Ubuntu 24.04 LTS
Minimum
2 GB RAM · 20 GB disk
Recommended plan
S2 (2 vCPU · 4 GB RAM · 60 GB NVMe)
Ready after boot
2 to 5 minutes (the gateway starts slowly on a small server)
Image slug
openclaw

About OpenClaw

OpenClaw is an open-source personal AI assistant that you run yourself. Its gateway talks to the language model you choose and serves a web Control UI to work with the assistant. Self-hosting keeps the assistant and its configuration on a server you control, and you pay the model provider directly for what you use.

This image installs Node.js 24 and OpenClaw from npm, with Caddy in front for HTTPS. On first boot the gateway is bound to localhost with password authentication, and a small setup page at /setup lets you connect a model provider with your API key and approve your browsers. You never need SSH to get started.

What’s included

Software installed on the image and the ports it uses.

Software
ComponentVersion
Ubuntu24.04 LTS
OpenClaw (npm package openclaw)Latest release at image build time
Node.js24 (NodeSource)
Caddy2 (automatic HTTPS)
Setup pageSmall Python service at /setup
Ports
PortProtocolUsed for
22TCPSSH
80TCPHTTP (redirects to HTTPS, certificate validation)
443TCP + UDPHTTPS: OpenClaw Control UI and the /setup page

Who it’s for

  • People who want a personal AI assistant on a server they control
  • Developers who want to try OpenClaw without installing Node.js and a reverse proxy themselves
  • Teams in the EU that want to choose their own model provider and keep the assistant's configuration on their own server

Use cases

  • A private assistant in your browser

    Open the OpenClaw Control UI at your server's HTTPS address and sign in with the gateway password.

  • Your choice of model

    Connect Anthropic (Claude), OpenAI (GPT), Google Gemini or OpenRouter with your own API key, and switch provider at any time on the setup page.

  • Room to build on

    It is your server with root access: configure OpenClaw further with its own CLI and documentation.

Resources

OpenClaw needs at least 2 GB of RAM and a 20 GB disk. Every plan includes an NVMe disk and 20 TB of traffic.

Smallest plan

S1

  • 1 vCPU (shared)
  • 2 GB RAM
  • 40 GB NVMe disk

The smallest plan that meets the minimum.

Recommended

S2

  • 2 vCPU (shared)
  • 4 GB RAM
  • 60 GB NVMe disk

A comfortable size for regular use.

OpenClaw needs at least 2 GB of RAM. The model runs at your provider, not on the server, so S1 is enough to start; S2 starts the gateway faster and leaves headroom for anything else you set up on the server. You can resize the server later. See current prices.

Getting started after deploy

  1. Wait for the first start

    On first boot the server sets up the gateway, the setup page and HTTPS. Until the gateway is up the address may show a 502 page. The first-boot log is in /var/log/nexoran-app.log.

  2. Find your address and password

    The address, user name and password are written to /root/APP-INFO.txt and printed when you log in over SSH as root. If you created the server with a root password and it meets the gateway's rules (any non-empty password), the app uses that same password, so you can sign in with what the console showed you at creation. If you used an SSH key (or the password does not meet the policy), a strong password is generated for this server and written to /root/APP-INFO.txt.

    ssh [email protected]
cat /root/APP-INFO.txt

  3. Connect an AI model

    Open https://<your-host>/setup, enter the server password, choose a provider (Anthropic, OpenAI, Google Gemini or OpenRouter), paste its API key and press Save and connect. The page runs OpenClaw's onboarding, restarts the gateway and shows the connected model. Until you add your own domain, the app is served at https://<your-ip-with-dashes>.sslip.io (for example https://203-0-113-10.sslip.io), a public wildcard DNS name that resolves to your server, with a Let's Encrypt certificate obtained automatically. Opening the bare IP address redirects there.

  4. Open OpenClaw

    Open https://<your-host>, enter the password in the Gateway secret field and press Connect.

  5. Approve your browser

    The first time a browser connects, OpenClaw shows "Approve this browser". Go back to /setup, enter the server password under Approve a browser and press Approve waiting browsers; the OpenClaw page then connects by itself.

  6. Use your own domain

    Create an A record for your domain that points to the server's IPv4 address and wait until it resolves. Then run the domain helper as root: it switches the app to the new name and obtains a new certificate. For OpenClaw it also updates the gateway's public origin in /root/.openclaw/openclaw.json and restarts the gateway.

    nexoran-app-domain app.example.com

Security notes

  • The gateway listens only on localhost (port 18789) and the setup page on localhost port 18790; Caddy is the only service reachable from the internet.
  • The gateway uses password authentication. The password is stored in /root/.openclaw/openclaw.json (readable by root only) and also protects the setup page, which refuses further attempts for 10 minutes after 10 wrong passwords.
  • A browser must be approved on the setup page the first time it connects, even with the right password.
  • The setup page passes your API key to OpenClaw's onboarding and does not repeat it in error messages or logs. Usage on that key is billed by your provider, so set a spending limit there.
  • OpenClaw runs as root on this server, so whatever the assistant is allowed to do, it does with root rights. Use the server for OpenClaw only.
  • A server without a firewall accepts all inbound traffic. Create an Odinex firewall that allows only SSH (22), HTTP (80) and HTTPS (443) and apply it to the server. Odinex firewalls filter traffic before it reaches the server, so Docker's own iptables rules cannot bypass them.
  • Take a snapshot before every upgrade, and turn on daily backups (20% of the server price) for anything you would not want to rebuild by hand.

OpenClaw questions

Clear answers about location, infrastructure and billing.

Which AI providers can I connect?
The setup page supports Anthropic (Claude), OpenAI (GPT), Google Gemini and OpenRouter. You need an API key from the provider. The connected model is shown on /setup, where you can also switch provider.
Is AI usage included in the price?
No. Odinex charges only for the server. Requests to the model are billed by your provider on your own API key.
Do I need SSH?
No. Connecting a model and approving browsers both happen on the /setup page. SSH as root is still available if you want to change more.
Where are the logs?
Run journalctl -u openclaw-gateway for the gateway and journalctl -u openclaw-setup for the setup page; systemctl status openclaw-gateway shows whether it is running. The first-boot log is /var/log/nexoran-app.log.
How do I update OpenClaw?
OpenClaw is installed globally with npm. Take a snapshot, then run npm install -g openclaw@latest and systemctl restart openclaw-gateway. Read OpenClaw's release notes before you upgrade.
Can I set up messaging channels or skills?
The setup page only connects a model; it skips OpenClaw's channel, skill and search setup. Configure those with the openclaw CLI as described in OpenClaw's documentation. Keep the gateway settings in /root/.openclaw/openclaw.json (localhost, port 18789, password authentication) as they are: the HTTPS setup relies on them.

Documentation

OpenClaw

Odinex Cloud

More from the marketplace

Deploy OpenClaw in Frankfurt

Your own server with full root access, NVMe storage and 20 TB of traffic, billed by the hour.

Deploy OpenClawAll apps