Your thoughts, yours.

Eidola is a personal AI with boundaries you don't have to take on faith. Every guarantee is verifiable in the architecture, not just promised in a policy. That's what frees you to apply it to everything you see, hear, think, and do.

First principles

Agency as architecture, not policy

Confidential by construction

On-device code mediates every request. When a task reaches beyond your machine, it runs inside secure enclaves — isolated environments that even the operators can't observe or modify. Open-source code, reproducible builds, and remote attestation make every guarantee verifiable, down to the unlinkable credentials leaving your device.

An extension, not a service

Eidola isn't a platform you send your data to. It's a tool that works on your terms and exists to extend what you can do — not to build a profile for somebody else.

Yours to keep

Your screen, conversations, documents, and patterns are processed into understanding that remains yours — not because we promise not to look, but because the architecture doesn't give anybody else a way in.

Why

Intelligence should be intimate

The most useful AI is the one that knows you best — your habits, your work, your conversations, the pattern of your days. That kind of knowledge is powerful and profoundly personal. It should be under your control, and every claim around it should be one you can verify yourself.

Most AI tools require you to trade this intimacy for capability. Eidola takes a different approach: we remove ourselves from the equation. Everything remains exclusively under your control, on your device or remote systems you can verify. It's not a policy you take on faith — it's a property of the architecture you can inspect for yourself.

In early development

When there's something to show, you'll be able to read every line, verify every build, and decide for yourself.