
Fantom Core
The module every Fantom is built from. One Fantom Core carries one session: a person or an AI agent at work.
Keys go in as light. Only video comes back.

Keyboard and mouse go from your side through a buffer, a beam of light across the air gap and an FPGA into the closed network. Video of the network's screen travels the other way, back to you.
A Fantom Core is two computers on one board, with an air gap between them. Your keyboard and mouse cross it one way, through a buffer, a beam of light and an FPGA, in a row. The screen comes back as video. No IP packet crosses, in either direction.
An AI agent works through a Core the way a person does: it reads the screen and types. There is no API into your network and no integration to secure.

One sealed module. One session.
Nothing is shared between one Core and the next: each session has its own path. Nothing else fits through it: no files, no network traffic.
What a Fantom Core is made of.
Two computers on one board, between the two halves of its heatsink.

Every Fantom is built from them.
Specifications
- Sessions
- One per Fantom Core
- Inside
- Two computers on one board, an air gap between them
- Into the network
- Keyboard and mouse, one way: a buffer, a beam of light, an FPGA
- Back from the network
- The screen, as video
- Across the air gap
- No IP packet crosses, in either direction
- Shared with other Cores
- Nothing
Tell us about your network and the agentic work challenges you face.

Next product: Zeroport Agent
Sovereign AI, on-device, judging every action in real time and physically terminating it when intent turns.



