Proof you can
run yourself.
Three tests we'll run with you at every demo, and that you can repeat at your own desk. No slides, no promises.
Pull the cable
Unplug the internet. Ask about your own files and last week's conversations.
What you'll see: it still answers. Your memory and your documents are on the box, so the internet isn't needed to know you.
Switch brains
Ask something that needs your history. Switch from OpenAI Codex to OpenRouter. Ask again.
What you'll see: same memory, same context, a different brain. Nothing had to be exported or re-taught.
Watch the wire
Open your router's traffic view, or any network monitor, and ask a hard question.
What you'll see: one masked request to the provider you chose. Our design goal is that nothing goes to inteligentia.
Watch it, unedited
Short, uncut recordings of the real box doing each test. We are filming them on production hardware; each one will appear here, with the date it was recorded.
Spec sheet
| Storage | 1 TB (€1,299) or 2 TB (€1,499) |
|---|---|
| Your agent | Built on Hermes Agent (open source), set up and tuned by us |
| On the device | Local language model, search over your files, long-term memory, chat history |
| Cloud brains | OpenAI Codex or OpenRouter, on your own account. Switch anytime. |
| Exposure filter | Masks names, IDs, account and phone numbers, emails and addresses before anything leaves |
| Talk to it | Telegram or another chat app you prefer, on your phone or laptop |
| Your past | Imports your chat history exported from OpenAI (ChatGPT), Anthropic (Claude) and others |
| Setup | Plug in power and network, pick a cloud brain, open your chat app. No terminal. |
| Case | Blue anodized aluminium, hexagonal ventilation |
| Status | Reservations open for the first production run. Specs may be refined before shipping. |
Straight answers
Is it 100% private?
No, and we'd be lying if we said so. Your memory, chats and files stay on the box. When a task needs a frontier model, a filtered slice of that request goes to the provider you chose. Here's exactly what leaves.
Do I need a cloud subscription?
For frontier-level answers, yes: your own OpenAI (Codex) or OpenRouter account. The box also works on its own for everyday things like finding, remembering and reminding, with its on-device model.
How is this different from ChatGPT, Grok or Claude?
Those are brilliant brains, and you can still use them, through the box. The difference is who keeps the mind: with them, the memory of you lives on their servers and works only with their model. With inteligentia it lives on your desk and works with any of them.
Can I bring my ChatGPT and Claude history?
Yes. Export your conversations from OpenAI or Anthropic (both offer a download), hand the file to your agent, and it imports them. Years of context, moved onto your desk.
How do I talk to it?
On Telegram, or another chat app you prefer, on your phone or laptop. Your agent is a contact; messages go to the box on your desk.
Couldn't I build this myself?
If you enjoy compiling drivers, wiring API keys and babysitting a server, absolutely, and the open-source core is there for you. inteligentia is for people who'd rather plug in a beautiful box and start talking.
What if inteligentia disappears?
Your memory and files are on a drive you own, and the agent at its core is open source. You're not locked in to the cloud vendors, and you're not locked in to us.
When does it ship?
Reservations are open now for the first production run. People on the list hear first, personally.