AI companion, not a therapist
Say what is on your mind, at any hour, in English, Hindi or Hinglish. If you are ever in danger, real people are one tap away.A private place to say what is on your mind, at any hour, in English, Hindi or Hinglish. SATHI is an AI, not a person. It listens, it stays with you, and if you are ever in danger it puts real people one tap away.
Beta for adults in India. Free. No account needed.
Some nights the hardest part is having no one to say it to. SATHI is a quiet light that listens, in your words, at your pace, and knows when a real person should be the one to answer.
How SATHI listens
Nothing clever is hidden. This is what happens between you pressing send and SATHI's first word.
Type in any language or mix. SATHI reads how you seem in this message: raw, anxious, venting, joking, hopeful. It reads it fresh every time.
A multilingual phrase list and an AI classifier look at the message in context. If you may be in danger, real people's numbers come before anything else.
Word by word, in your language and your register, with one question at most. In a crisis there is no AI text at all: one human-written line and the numbers.
What SATHI does
Say as much or as little as you like. Replies stream in word by word and follow your tone, not a script.
work has been really heavy this week
Heavy in the piling-up way, or heavy in the slow, grinding way where even small things cost something?
the second one honestly
That kind wears you down quietly. What has been asking the most of you?
Write the way you speak. SATHI replies in the same language and script, slang and all. Tap a language.
I don't feel like myself lately
Not like yourself in a flat, far-away way, or restless and on edge?
No name, email or phone number. Encrypted before it is stored, in Mumbai.
One conversation, your profile, or everything. Immediate and permanent.
Tele-MANAS and 112 from every main screen, including chat. Works offline.
SATHI breathes, blinks and glances with you. Press and hold it and it breathes with you.
A calm sage paper by day, deep ink by night. It follows your phone, or your choice. Tap to switch.
Write in any of these. English, Hindi, Hinglish and Spanish are safety-tested; other languages are translated for the safety check first.
Safety promise
These lines are free, confidential and open to everyone in India. SATHI cannot call for you; you make the call, and the numbers are always one tap away.
Every message is checked before SATHI replies. A message that may mean danger, to you or to someone else, gets one fixed, human-written line in your language and the numbers above, at once, with no animation and no waiting for the AI. For a medical emergency, 112 comes first.
Get support stays pinned in the header for the rest of that conversation. If you say you were joking, SATHI does not argue and the card stays. Under-18 disclosures get a kind stop with support lines, and the conversation is deleted.
Privacy in plain words
The short version of the Privacy Notice, in the same words.
Meet SATHI
SATHI is a point of light that grew eyes. It has no age and no backstory, and it never pretends to be human.
On your phone
Taken from the app this week, not mock-ups. The test crisis message is one of the phrases from our safety set.
The first screen: an AI companion, not a person and not a therapist. Private, deletable, and real people one tap away.
Replies stream in word by word and follow how you seem in this message, with one question at most.
A message that may mean danger gets a fixed, human-written line and India's crisis lines at once. No AI text.
Tele-MANAS and 112 from every main screen, including the chat header, with or without a connection.
About me shows what SATHI knows. Change the look, or delete a conversation, your profile or everything.
Get SATHI
Two routes to the same app. Both are free; both ask for nothing but your year of birth and consent.
A direct install (APK), about 148 MB. Android 7.0 and up.
The same app in Safari, added to your Home Screen. Nothing from the App Store.
Questions
No. SATHI is an AI companion. It says so on every screen, and it is not a replacement for a therapist, a doctor or a friend. It is a private place to think out loud, and a quick route to real people when it matters.
Every message passes a multilingual phrase list and an AI safety check. If you may be in danger, SATHI shows one fixed, human-written line and India's crisis lines at once (Tele-MANAS 14416 and 1-800-891-4416, Emergency 112), with 112 first for a medical emergency. Get support stays pinned for the rest of that conversation. SATHI cannot call anyone for you.
You. Everything is encrypted before it is stored and only the running app holds the key. Nobody on the SATHI team reads conversations unless you opt in to de-identified research review. During the beta your messages are also processed by our AI providers; see the Privacy Notice for exactly what that means.
Replies come from Google Gemini (free tier: Google may use content to improve its products and human reviewers may read it) or, when Gemini is busy, from Groq (which does not retain prompts by default). The safety check runs on Groq first, then Gemini. The crisis line itself is never written by an AI.
No. There is no login, email or phone number. SATHI asks for your year of birth, because it is for adults 18 and over, and for your consent. That is all.
This is a closed beta. The APK installs directly from sathi.world; Android asks once to allow installs from your browser. On iPhone, SATHI runs as the same app in Safari and can be added to your Home Screen.
The crisis lines, the safety wording and the legal checks are done for India first. From outside India the app shows a not-available screen with a worldwide helpline directory.
It can be. The beta runs on free AI tiers, so at peak a reply may take longer or SATHI may say it could not reply just now, with Try again. The crisis path never depends on the AI.
Yes, in About me: Delete everything removes your account, profile, conversations and counters immediately and permanently, and takes you back to the start.
Not yet. A clinical psychologist will review the safety rules and crisis lines before the wider beta. The Safety Protocol and Privacy Notice are drafts, and we say so.