Your conversations stay private.
No conversations stored. Hosted in Switzerland. Automatic anonymization.
Start a conversation
Try instantly · No sign-up
Truly personal AI.
No trade-offs.
Privacy isn't a constraint here, it's the foundation. Because nothing leaves your device, you can finally ask the real question — at work as in life — and point the most powerful models at it.
-
Local-first: everything stays with you
Your conversations and your personal memory live in your browser and nowhere else. Nothing is stored on our side: even if we were asked, we would have nothing to hand over.
-
Your personal memory, in plain sight
Your peanut allergy, the treatment you're on, the project's real budget. What you'd never type anywhere else, the AI remembers here — and you fix it in one click.
-
Open source, hosted in Switzerland
Open models, auditable by anyone, on Swiss servers. The country that guards data the way it guards everything else.
-
One model per question, not per conversation
A quick question doesn't need a model that thinks for three minutes. A careful analysis does. Let Auto mode handle it: it switches on its own, message after message.
-
Got a favourite external model?
Keep it. It receives your question anonymised — anything identifying you is replaced before sending, your profile never leaves, and you have the final say.
Anonymised on the way out. Enriched on the way back.
The best models on the market aren't all open source. You still get their power: our open models frame them, on the way out and on the way back.
ChatGPT, Claude, Gemini, DeepSeek
All reachable from the same interface. No account with each of them, no history left behind.
-
What identifies you is spotted
A local analysis combs through your question: names, addresses, amounts — and anything that would make you recognisable in passing.
-
You see it, you decide
The masked version appears, editable. And if your data is too identifying to be masked reliably, nothing goes out: we stay in Switzerland.
-
The external model answers a stranger
It works on the masked version. Neither your profile nor your memory is passed on: it has no way of knowing who is asking.
-
Personalised here, not there
An open source model takes its answer and reworks it with your full profile and memory — which never left. That's the version you read.
Your own rules
You write your own substitutions: your name, your company, your clients. They apply to every external request.
This isn't an option to switch on: it's the architecture. Included in every plan.
Flat rate. Not a meter.
Elsewhere you may be sold usage-based consumption.
You watch your tokens instead of moving forward.
-
Several AIs, one app
Switch models whenever you like, or let Auto mode pick the most relevant one for each question.
-
Routing is what makes the flat rate possible
Each request goes to the model — or the service — that suits it. No need for a datacentre to give today's weather. What we don't spend, you don't pay for.
-
Our standard models, always unlimited
External and specialized model usage is diluted into our standard models in Switzerland. The experience stays smooth.
A context that stays light
We automatically keep your context lean: with every question, only recent exchanges and a summary of the earlier ones are sent to the model.
Privacy by design, whatever your usage.
- Plus 19 € / month Optimized usage · automatic routing
- Business 49 € / month Intensive usage · configurable routing Coming soon