Create your first desktop
Step by step: create your first SpineOS cloud desktop, pick a plan and a model, and watch it boot with your agent on board.
Before you start
You will need:
- A SpineOS account (see Create an account)
- An account or API key for your agent: a Claude, ChatGPT or xAI account, or an API key from Anthropic, OpenAI, or OpenRouter (see Configure AI provider)
Step by step
1. Open the creation form
From your dashboard, click "New desktop". A modal opens where you configure your desktop.
2. Pick a name
Give your desktop a name you will recognize later, like "research-agent" or "code-assistant". It shows up in your dashboard.
3. Select a tier
Choose resources based on what your agent needs:
- Free -- 2 vCPU, 4 GB RAM, 20 GB disk. 10 tasks/day. Good for trying things out.
- Pro -- 4 vCPU, 8 GB RAM, 40 GB disk. Unlimited tasks. For real work. (EUR 49/mo)
- Max -- 4 vCPU, 16 GB RAM, 80 GB disk. Unlimited tasks. For heavy workloads. (EUR 149/mo)
Every tier runs on your own AI account -- connect a Claude or ChatGPT subscription, or bring your own API key (BYOK).
You can upgrade later.
4. Choose your agent type and AI provider
Pick an agent type:
- OpenClaw -- general-purpose agent. Works with any model via Anthropic, OpenAI, or OpenRouter.
- Claude Code -- Anthropic's coding agent.
- Hermes Agent -- Nous Research's self-improving agent.
- Codex -- OpenAI's coding agent.
- Grok -- xAI's coding agent, Grok Build.
Then give the agent what it runs on, in the same form: an account you connected in Settings, or a provider API key you paste (Anthropic, OpenAI, or OpenRouter), as the form offers for that agent. Grok has nothing to set here: it signs in with your xAI account on the desktop once it starts. When you paste a key, pick a model from the model selector, which searches available models from your provider's API.
You can also add an embeddings API key to give your agent persistent memory across conversations. This is optional.
5. Create
Click "Create desktop". SpineOS will:
- Clone a pre-built desktop image
- Inject your configuration
- Boot the VM
- Start the AI agent and supporting services
What happens during provisioning
Your desktop goes through two states:
- Booting -- the VM is starting, initialization is running, services are being configured. Usually 30 to 60 seconds.
- Running -- ready to use. VNC viewer, terminal, and chat panel are all available.
The status updates in real-time on your dashboard. Once it shows "Running", click the desktop to connect.
Troubleshooting
Stuck on "Booting"? A new desktop is usually working, agent included, within about 2 minutes. If it is still on "Booting" after 3 minutes, refresh the page. If the status still does not change, see Desktop won't start.
"Create desktop" stays disabled? If the form says "This desktop needs your own AI credential.", the agent you picked has no account or API key yet. Connect an account or paste a key, as the form offers under that message.
Next steps
Desktop running? See Connect to desktop for how to use the VNC viewer, terminal, and chat panel.
Related docs
Configure AI provider
Set up your own Anthropic, OpenAI, or OpenRouter API key on SpineOS so your agent runs on the model and the account you choose.
Connect to your desktop
The three ways into a SpineOS desktop: the live screen in your browser, the integrated terminal, and the chat panel your agent works in.
Understanding tiers
What each SpineOS plan gives you in processor cores, memory, storage, and desktop count, and how to tell which one you need.