Pick an AI agent.
Start in one click.
OpenClaw, Hermes, and more — ready to use in your browser. No local installation. No setup. Keep AI off your computer. Keep it running 24/7.
Four steps from image to running workspace
Every container is created through the same short form. Select a step to see what it asks for.
See the price before you launch
Each resource tier bills by the hour in credits, covering compute and the base volume. Nothing is charged for compute while a container is stopped.
| Resource | vCPU | Memory | GPU | Base volume | Credits / hr |
|---|---|---|---|---|---|
| cpu-2-memory-2Gi | 2 | 2 GiB | — | 60 GiB | 30.12 |
| cpu-4-memory-8Gi | 4 | 8 GiB | — | 60 GiB | 72.12 |
| cpu-8-memory-16Gi | 8 | 16 GiB | — | 60 GiB | 144.12 |
| cpu-32-memory-32Gi | 32 | 32 GiB | — | 60 GiB | 1,000.12 |
| cpu-8-memory-16Gi-gpu-RTX3090-1 | 8 | 16 GiB | 1 × RTX 3090 | 60 GiB | 544.12 |
| cpu-8-memory-16Gi-gpu-RTX4090-1 | 8 | 16 GiB | 1 × RTX 4090 | 60 GiB | 744.12 |
| cpu-12-memory-32Gi-gpu-RTX4090-1 | 12 | 32 GiB | 1 × RTX 4090 | 60 GiB | 840.12 |
| cpu-32-memory-64Gi-nfs-128-gpu-RTX4090-2 | 32 | 64 GiB | 2 × RTX 4090 | 128 GiB | 1,792 |
What you get with every workspace
An image that ships VS Code or Jupyter has it published as an endpoint the moment the container is running — no local setup, no port forwarding.
Reach the same workspace from any machine over the browser. No SSH keys to distribute, no local CUDA to install.
Register any public or private registry image. Private pulls use an ImagePullSecret you store once and reuse.
Upload a dataset once, then mount it into any container in the group instead of copying it per workspace.
Images, datasets and containers belong to a group. Invite the members who need them and share access explicitly.
Balance and per-container usage are visible on the billing page, so a group can see where its credits went.
Sign in and start your first container
Accounts are issued through your organisation's login. Once you are in, creating a workspace takes one form and a few minutes.