Free. €29. €49. Custom.

Depends whose key runs the agent, and we would rather say so than sell you a slogan. Add your own key — a free Gemini one takes 30 seconds — and there is no platform-side limit: the cap is your provider's, they set and change it, and the live numbers are in your own AI Studio dashboard. When you hit them the agent waits, or routes to your paid key if you added one. Run without a key and your agent uses our included model (Azure OpenAI, on our account), which we keep an internal spending cap on the free track (we do not publish the figure) — because that bill is ours, not yours.

My time, asynchronously. You say "I want X". I reply in 24h on business days with either a working setup, an explanation, or a question. If the agent cannot learn something alone, I write the custom script. Same async rules — no calls.

€49 one-off. I install the full OpenClaw agent on hardware or a VPS you already own, buy, or rent separately. We pick the models together (Gemini free / OpenAI / Claude — your keys), wire a Telegram or Slack bot, and set 2-3 starter automations. You get 7 days of written support via the Telegram support bot — anything breaks in week 1, I fix it. After 7 days you can add Light maintenance (€9/mo — I watch the heartbeat and restart on crash) or keep maintaining it yourself with the PDF guide. Prefer to try first? Hosted Auto is free.

While your agent is awake, yes. Free agents sleep after 3 quiet days — we host them for free, so idle boxes rest — and scheduled tasks pause with them. Use your agent at least once every 3 days (the dashboard chat or your Telegram bot both count) and everything keeps running. We email you a day before it dozes off, nothing is ever lost, and waking is one click. Paid plans never sleep.

One click in dashboard → a tarball of your agent's whole state: config and its backups, workspace, the skills you wrote, memory, and its databases — scheduled tasks in state/, and history and embeddings under agents/. Restore it onto your own OpenClaw install — it is the agent's state, not a boxed appliance. A few things sit outside it (raw session transcripts, your account records) and we send those on request; /gdpr lists them. We hate lock-in more than you do.

Almost. ChatGPT Plus and Claude.ai are chat apps — they don't include API access. To plug into the agent you need an API key: platform.openai.com/api-keys or console.anthropic.com. Those are pay-per-token, separate billing from any subscription you might already have. We're not in the middle of it.

Yes. There's a web terminal in dashboard. Add cron jobs, install packages, peek at logs. If you break it on free Hosted Auto, restoring is on you — that's the deal. On Assisted (€29), I'll un-break it.

ChatGPT waits for you to open a tab and ask. Your agent works on its own: it runs crons, watches sites and prices, triages messages, and reports to your Telegram — around the clock, whether you're online or not. You can still chat with it any time; the difference is it also acts when you don't.

Two places, honestly stated. Your agent's setup, config, logs and encrypted keys live on our EU server (Hetzner, Germany). When the agent needs a model answer and you have added your own key, the prompt goes to YOUR provider on YOUR key — Google Gemini if that is what you chose, under their terms, possibly outside the EU. If you have added no key, it goes to our included model instead: Azure OpenAI on our account, so your prompt passes through us. Our Azure resource sits in the EU (Sweden Central) and stores data at rest there, but it runs on Azure's global standard deployment, so the inference itself may happen outside the EU — the same caveat we give for your own provider. We never resell your data or train on it, and you can export or delete everything in one click.