Note: This service is provided for experimental purposes only and is offered "as is". We have no control over or liability for their behavior.
Performance and availability are not guaranteed, especially during periods of high demand. No backups are created, so you are responsible for backing up your own configurations and should not use the platform for production environments.
Overview
OpenClaw is the first software being provided via the Experiments Platform. Your OpenClaw instance runs in a container that comes preconfigured for easy and secure use. You can use it to simply chat with some LLMs hosted by us, or to set up some advanced, personal-assistant-like automation.
Getting started
Click Create OpenClaw Instance
Under Additional Settings, leave all checkboxes as they are.
The size Shrimp should do fine to play around at first. You can rescale your instance to a larger size later if needed.
Provide a UI password or hit Generate a password. Save this password, you'll need it later to access your OpenClaw instance's web interface, using the username openclaw.
Provide a name for your OpenClaw instance. This name is purely cosmetic for displaying it in the UI.
Wait about one minute until your instance is ready
Wait for the little status LED in front of your instance to turn green.
Visit your OpenClaw instance's UI.
You can find the URL in the experiments web interface. Your instance is publicly reachable, so we secured it using the password you provided earlier. To log in, use username openclaw, and the password you provided earlier.
Start chatting
Please be patient, the first request of a new session might take a little bit longer than usual. Each request after the first will be much faster. Give it a try!
Advanced configuration
To look into more advanced configuration options, check out OpenClaw's own documentation.
FAQ
What makes Hetzner's OpenClaw instances special?
Our OpenClaw comes ready for use, with a hardened configuration. It also uses Hetzner-hosted LLMs out of the box. To achieve this, we do not allow direct access to the instance's file system. You can still use all of OpenClaw's usual features.
What level of stability and reliability can I expect?
This is an experimental product. As such, some bugs should be expected. Data loss can not be ruled out. It is your responsibility to back up your data.
The Hetzner-hosted LLMs are also provided on a best-effort basis. While we make use of tools such as rate-limiting, they might still become temporarily unavailable during times of high demand.
Which AI is OpenClaw connected to?
By default, OpenClaw is configured to use a selection of LLMs hosted by Hetzner. The UI's chat window allows you to switch between the different models (At the bottom, click on Qwen/Qwen3.6-35B-A3B-FP8). If you have an account for an externally hosted AI provider such as Claude or ChatGPT for example, you can configure those as well.
Which of the internal models should I use for which purpose?
None of the models are strictly "better" than the others. Feel free to experiment, and let us know what you find out!
How do I use the openclaw-cli without access to the file system?
The openclaw-cli commands can be useful for managing your OpenClaw instance. Usually, you would run these commands directly on the OpenClaw server. For security reasons, our OpenClaw does not allow out-of-band shell access.
But there is an easy workaround: You can run any OpenClaw commands directly in the "Chat" tab of your instances's web interface: simply type the OpenClaw command directly in the chat.
For example, typing openclaw status will run the status subcommand of the openclaw-cli.
What do the two checkboxes under "Instance Settings" actually do?
-
"Reset config defaults on Startup": this makes sure a few config settings that are required to make our OpenClaw work are merged in with your own config at startup. This includes basic config to make connectivity of the UI work as well as secure defaults and LLM configuration. This only overrides conflicting options that can not be merged.
-
"Run OpenClaw Doctor on Startup": this is only meant to troubleshoot a broken OpenClaw instance that you cannot recover otherwise
Where should I direct questions and feedback?
If you have a specific problem or question, you can also contact us via the Support Form in the Experiments UI. For general questions and feedback, we'd be happy to hear from you in the Forum.
Are changes to my OpenClaw instance persistent?
Each instance of OpenClaw has some storage that is used to save configuration and context between container restarts. That storage is mounted at /data and replicated to avoid data loss in case of hardware failure. However, once you delete an instance, all associated storage is permanently discarded.
Can I create a backup of my OpenClaw instance?
We do not offer a backup functionality at this time. You can retrieve much of the configuration info from your OpenClaw's UI, including its "personality settings" such as AGENTS.md or SOUL.md. Some sensitive information is hidden in the UI, however.
You can however also run commands via the "Chat" functionality of your OpenClaw's web interface to back up its configuration and state. If your chat message starts with run, you can tell OpenClaw to execute a command.
For example, run cat /data/openclaw.json will display the configuration. This includes sensitive data, so take care. Other configuration options, such as your AGENTS.md or SOUL.md can be retrieved via the configuration UI of your instance.