What is Sathi?

Sixty seconds

Sathi is a personal AI built to grow around your life. It connects what a life is actually made of — conversations, documents, mail, calendar, people, decisions — into one continuous understanding that belongs to you, so you can ask the questions a household actually asks: when does Mom’s flight land, what did we decide about the house, tell Seerat I’ll be late. It is one intelligence rather than a per-device assistant: the same Sathi appears in the app, in mail, on a speaker, in the home — and carries the same understanding onto every surface it comes to inhabit. It speaks Nepali, English, Japanese, and Hindi. It is built by Sathi Systems on the evolving Altruistic AI architecture, and is the living implementation through which that architecture is being learned. It is in beta: one production deployment, its first families in daily use, no billing today, signups open for a testing window, and guest chat open to everyone — nothing becomes memory; guest conversations are kept briefly for testing.

The three names

One sentence relates them, the same everywhere: Sathi is built by Sathi Systems on the evolving Altruistic AI architecture, and is the living implementation through which that architecture is being learned.

Who it is for

Individuals and their households, first. One Sathi per person; a family shares context only where its members choose to — permission-structured from day one, private by default. The household is not a marketing angle: it is the first proving ground, because a family is the smallest group where continuity, shared memory, and boundaries all have to work at once. From there, trust opens outward in rings — builder → household → kin → communities → anyone — and each ring earns its opening only when the people in the previous one genuinely prefer Sathi to the incumbents.

What it can do today

Live in daily use, as of 2026-09-05: Memory & understanding · Conversation · Documents (vault) · File authoring · Calendar · Reminders, tasks & alarms · Email · Family messaging & calls · Contacts & relationships · Web search & reading · Notes & lists · Music & photos · Learn · Archive imports · Export & exit.

Experimental — built and real, but dark, flagged, or prototype-grade: Voice · Devices & home.

Everything else is labeled planned or conceptual and stated as such — nothing unbuilt is claimed as fact. The full inventory, with status, actions, authorization, and the data each capability touches, is /capabilities (machine twin: /capabilities.json).

Why it is different

The thesis is continuity, and privacy is what makes the continuity trustworthy:

• One intelligence that keeps knowing you. Understanding is built from your own archive as claims that carry their source, their time, and their confidence. It compounds across years, and survives model swaps, device swaps, and vendor changes — the relationship is the product, not the session.* Every surface is an embodiment of the same Sathi, never a separate assistant with its own memory.
• Ownership is what makes being known safe. You and your family own your data and the understanding built from it: total signed export, provable deletion, a front door that accepts your archive back.* An AI this close to your life is only tolerable if leaving is always mechanical, never a negotiation.
• Models are replaceable; the relationship is not. Underneath, models and agent runtimes are interchangeable parts behind seams. Your understanding never belongs to whichever brain is currently strongest.*

The honest comparison with generic assistants — including what they do better today — is /why-sathi.

What stage it is at

• Beta. One production deployment; its first ring of real users — the builder’s own family — in daily use, in all four languages.
• Signups: open for a testing-and-feedback window (the gate is a runtime knob the operator can close when the window ends).
• Guest chat: open to everyone, no account; nothing becomes memory, and guest conversations are kept only briefly for testing.
• Pricing: access is free during the current beta; no billing exists. Future hosted plans will publish finite cloud allowances before anyone pays.
• Hardware: a home speaker on open hardware is experimental; nothing is manufactured or sold.
• Openness: the Altruistic AI research notebook is public (v0.8); the reusable framework and Sathi’s own source code are not published today. Distillation and open publication follow saturation and repeated frontier-system, human, and security review.

Deployment model

Sathi runs today as a private deployment operated by Sathi Systems — your archive and understanding live with that Sathi, never on a model provider’s. Model providers see per-request context, not the store. The same Sathi is being shaped so each capability can run on the nearest capable and authorized node: home hardware, private infrastructure, Sathi infrastructure, or an external provider — without splitting identity or moving understanding merely because compute moved. Home/local is experimental; hybrid routing is direction; dedicated and fully independent deployments are future offerings. Organizational deployments are available for conversation on sathisystems.com; nothing is signed today. Intelligence may be rented; understanding stays home.

Questions, answered the same way everywhere

What is Sathi?

A personal AI built to grow around your life.

Who makes Sathi?

Sathi Systems K.K., operating across Japan and Nepal.

What is Altruistic AI?

A personal notebook asking what altruism means, with notes on intelligence, cooperation and systems. The generalized framework and reusable code are not released yet.

Why “Altruistic”?

Because we are interested in intelligence that serves, rather than intelligence that has to preserve, expand or advance itself.

Who is Sathi for?

Individuals and their households. One Sathi per person, shared family context by choice. Today it serves its first ring of real users in daily use, in Nepali, English, Japanese, and Hindi; trust opens outward in rings from there.

What does Sathi cost?

Access is free during today's beta because billing has not launched. Future hosted plans will publish finite cloud allowances. There is no advertising, no engagement optimization, and personal data is never sold.

What stage is Sathi at?

Beta. One production deployment serves its first families in daily use; signups are open for a testing-and-feedback window; guest chat is open to everyone, keeps no memory, and is logged only briefly for testing; a home speaker on open hardware is experimental.

Where everything else is

The site is five doors, one audience each (2026-09-01) — so a person deciding whether they want Sathi and an engineer reading how it is built are never handed the same page:

• Sathi — the front door: what it is, four things it does, the trust facts, the stage, and the way in.
• How it works — the plain-English explanation, in four languages.
• Trust — where your life is kept, what leaves, how to take it all with you, and the honest limits. Detail: Privacy and Security.
• Explore — the one technical entrance, and the only place that reads like documentation: architecture, memory, models, devices, capabilities, experiments, interfaces, implementation status, the honest comparison, and Principles — why it exists and what it refuses to do, in four languages.
• About — the company, the stage, who owns what, and how it sustains itself.

The framework Sathi implements is not re-taught anywhere on this site: its axioms, ontology and recursive composition are published in four languages at altruistic.ai.