What we can and cannot see, and why the architecture, not our word, is the answer.

Privacy model

What is true today. What is being built.

Your data lives on a dedicated server, isolated to your account and not shared with any other user. It is never used to train any model. Two things are true about it right now: only you can access it through Mandaire, and no one on the Mandaire team can read it through any normal operational channel (no admin panel, no support tool, no bulk access path).

We are building toward a stronger guarantee: zero-knowledge cryptographic primitives that will make it architecturally impossible for anyone, including us, to read your data even under legal compulsion. The goal is the same guarantee Signal has for your messages: a government request or hosting breach produces encrypted data the requester cannot read, because we do not hold the key. That architecture is under active development and not yet complete. We will update this page when it ships and what ships.

The LLM that reasons over your context is the one you bring: Claude Max, ChatGPT Plus, Gemini. Mandaire holds the context; the inference runs on your existing subscription. Neither sees more than the specific context needed for the specific question. No bulk archive is sent to any provider.

If Mandaire shuts down, you get your data back. Full export in portable formats. Mandaire is built on widely-audited open-source components and the architecture is fully documented. No lock-in is an architectural commitment, not a preference.