Mandaire is the private context layer underneath the AI you already use. It joins your email, messages, calendar, contacts, photos, notes, docs, files, and AI conversations into one source-backed picture, and decides what each assistant or app gets to see. The data is yours, the keys are yours, the server is yours. Keep using the walled gardens. Stop being stuck inside any one of them.
Mandaire comes from "mandatary," a legal term for a person authorized to act on behalf of another. That is the relationship: a digital entity that knows your full context, works exclusively for you, and acts only with your authorization. The logo is the same idea: two rectangles, one solid for you, one outlined for your digital counterpart. Human and machine. 1 and 0.
Mandaire is being built by a small team. The founder has thirty years of professional experience across consumer technology, marketplaces, strategy, and data science, much of it leading large analytics and data organizations. Identities are kept off the public site while the product is in private beta. Investors, journalists, and prospective users who want a direct line can request it at [email protected].
Building anonymously costs a trust signal that we do not take lightly. We compensate for it by publishing what can be checked now, not what could be checked later: the architecture and threat model, the written privacy and data-handling commitments, the connection mechanism and scopes for every supported source, the disclosure-engine policy primitives, and the data-export format. The encryption module is on track for open source and a third-party audit in Q3 2026. Identities will be named at general availability or earlier on direct request.
Mandaire runs on a Linux server you own, in your own cloud account or on your own hardware. You reach it three ways: over private chat (encrypted, just you and your data), through the AIs you already use (Claude, ChatGPT, Gemini, or any MCP-compatible client), and through the apps you've connected (email, messaging, calendar, notes, docs, files; Mandaire drafts replies, writes notes, and suggests responses, never sending without your sign-off). There is no native Mandaire app to install. Most providers (Google, Microsoft, Meta) connect over standard OAuth. Apple sources (iMessage, Photos, Notes, Calendar, Contacts) need a small local agent on a Mac you own, since Apple doesn't offer consumer OAuth for these.
Philosophy
The default arrangement on the consumer web is that the company you signed up with is the legal custodian of your information, and your access depends on their continuing goodwill. Mandaire inverts that. The cloud account is in your name. The server is in your name. The encryption key is in your possession. We don't have the bill, we don't have the key, and we never become the custodian of your data. If we disappeared tomorrow, you'd still have everything.
A promise not to look at your data is not the same as being unable to look at it. Mandaire is built so that your encryption keys never leave your control. The privacy guarantee comes from the architecture, not from anything we say.
Reading your data is low risk. Sending a message on your behalf is high risk. Mandaire starts as a read-only intelligence layer that advises and drafts. Write actions ship behind a draft-then-approve gate. Outgoing actions are earned over time and always remain under your control.
Most AI tools optimize for sounding confident. Mandaire optimizes for being right enough to act on. Every claim is sourced. When the corpus has a gap, Mandaire flags it instead of making something up. "No notes from your September dinner. Outcome unknown." beats a plausible-sounding fabrication every time.
The encryption module that handles your data is on track for release as open source with reproducible builds in Q3 2026, audited by an independent security firm on the same timeline. Until then, the architecture, threat model, schema, disclosure-engine primitives, and export format are published in plain text on this site so you can evaluate the running system today. We don't want you to take our claims on trust. We want you to be able to check.
People adopt AI thinking they want a faster assistant. What they actually need is a thinking partner that handles the digital overhead so the time they spend with people is higher quality. The shift from "do this for me" to "think about this with me" is when AI partnership starts to be worth something.