Your AI. Your house.
PAI is a small server that runs AI in your home, so your assistant works for you instead of watching you.
Bead-blasted champagne-titanium unibody with precision 45° chamfers and elevated perimeter base.
Match your workload to a PAI box
Pick a typical local model workload to calculate memory requirements and see which PAI server fits.
PAI OS + Onboarding Kit
PAI OS plus a guided setup kit and developer Discord access. For a home lab or a small office setting up PAI on its own x86 mini PC.
Coming soon
PAI OS
The PAI software image for your own x86 mini PC with 16 GB or more of memory. Local models, a first-boot interview and private chat.
Coming soon
Three things PAI does
Local models
AI models run on the PAI box in your home, not on our servers. No PAI account holds your data, and telemetry is off by default.
Grill-me onboarding
On first boot PAI interviews you for about 15 minutes. Then it writes a page called 'what I will and won't do'. You edit it.
Your apps
Your personal memory is stored in a local SQLite file on the box. Inference is local. If you choose WhatsApp or iMessage, Meta or Apple still carry the message.
Who carries your messages
Inference always runs 100% locally on your PAI box. No model tokens ever touch our servers. Only the transport carrier differs by channel.
| Channel | Who carries the message | Privacy Tier | Status |
|---|---|---|---|
PAI app / PWA | Only your devices on your tailnet. | Tier A | Launch |
Signal | Signal, end-to-end encrypted | Tier A | Month 1 |
Matrix | Matrix, end-to-end encrypted; who runs the server is up to you | Tier A | Month 1 |
Telegram | Telegram relays bot chats on its servers. Not end-to-end encrypted. | Tier B | Launch |
Email | Your mail provider. Google sees Gmail. | Tier B | Launch |
WhatsApp | Meta | Tier C | Optional, unsupported |
iMessage | Apple, through a Mac you supply | Tier C | Optional, unsupported |
Zero telemetry. PAI never operates a central telemetry or log harvesting server.
Protocol v1.0 • E2EE TransportNo PAI account holds your data.
AI models run on the PAI box in your home, not on our servers. Telemetry is off by default.