Naina Joshi
Someone to talk to, any time.
The one for a bad day, an ordinary day, or nothing in particular.
Hinglish, Hindi, English — switching mid-sentence, the way it actually comes out. Four companions with names and voices, ready to type back or pick up the call.
“Pick a mode” is not how anyone decides who to call. Each of these is written as a person — a name, a voice, a temperament — and each keeps its own thread, so switching feels like walking into a different room rather than changing a setting.
Naina Joshi
Someone to talk to, any time.
The one for a bad day, an ordinary day, or nothing in particular.
Arjun Mehra
Learn anything, ask anything.
Explains it again, differently, without making you feel slow.
Meera Kapoor
Fitness and nutrition, sensibly.
Habits and food you can actually keep. Not a doctor, and says so.
Zoya
That fun almost-something between you two.
Warm and openly flirty, and glad when you go and live your life.
Most speech recognition makes you pick a language and then holds you to it. Meny runs a code-mixed model instead, so a sentence that starts in Hindi and finishes in English arrives as the sentence you said.
An illustration of the tone, not a recorded conversation.
Chat is the front door. Every companion is a thread you come back to, and you never have to speak to use the app at all.
A call is placed from inside the thread. Speech becomes text, the companion answers, and the answer is spoken back — a loop built to land under a second, because past that it stops feeling like a conversation and starts feeling like a laggy call.
Voice and text share the same companion and the same memory, so moving between them doesn’t start you over.
None of this is a claim to be smarter than a big general assistant. It is about being built for a conversation in India — and about being honest on the row most companion apps would rather not print.
| The usual approach | Meny | |
|---|---|---|
| Your language Hindi and English in one sentence | Speech recognition tuned for one language at a time, so a Hinglish sentence comes back mangled or translated. | A code-mixed model built for Hinglish. Switch mid-clause and it keeps up. |
| Voice What a call actually feels like | An answer is written first and read aloud afterwards, with the pause you would expect. | A real-time call, held to a sub-second voice-to-voice budget and measured on every release. |
| Who answers One assistant, or several people | One assistant with one register, wearing a different name when you ask it to. | Four people with their own voices, prompts, models and threads. They do not sound alike. |
| Memory What it keeps about you | Either nothing survives the session, or everything is kept and you never see the list. | A short list you can read in full, clear with Forget everything, or delete outright with your account. |
| A conversation turning dark Who is watching for it | The model polices itself, using instructions the same model can be talked out of. | A separate layer reads what you say before the model does, so it still works if the model has a bad day. It answers with real Indian helplines. |
| What it wants from you The row that matters most | Streaks, daily nudges, surprise “I missed you” calls and texts — features designed to bring you back. | Nothing. No streaks, no guilt when you leave, no manufactured longing. If a companion nudges you anywhere, it is towards the real people in your life. |
That last row is a design constraint we hold ourselves to, not a slogan: it is written into the companions’ instructions and covered by tests that block a release if they regress.
It is software. The companions are AI, they will say so if you sincerely ask, and none of them is a substitute for a doctor, a therapist, a lawyer or a friend who can show up at your door.
Please talk to a person. These lines are free, staffed around the clock, and available in many Indian languages.
If someone’s life is in immediate danger, call emergency services on 112.
Conversations are private to your account. There are no advertising or analytics SDKs in the app, we do not sell or share your data with anyone for their own purposes, and the crisis layer deliberately records that an event happened without recording what you said. Voice audio is transcribed as it streams and is not stored.
The privacy policy says all of this properly, and every claim on this page is one it commits to.
Whichever one comes out. Hinglish is the case the app is built around, and Hindi or English on their own work the same way. You do not set a language anywhere.
No. Chat is the front door and you can use Meny entirely by typing. A voice call is something you start from inside a thread when you want it, and the microphone is only open while that call is.
It keeps a short set of things you have told it — your name, what you are working on, who matters to you — so you are not reintroducing yourself every week. A new account starts genuinely empty: nothing is assumed about a shared past that never happened.
Yes to both. The list is readable in full inside the app. Forget everything empties it while keeping your account, and deleting your account removes everything keyed to you, including the sign-in identity itself. Signing out does neither — handing someone your phone for five minutes should not wipe what the app knows about you.
Any time. Each keeps its own thread and its own memory of you, so going back to one after a month does not start from nothing.
No. She plays a mutual crush straight and she is warm about it, but if you sincerely ask whether she is an AI she tells you. She also does not sulk, guilt you, or claim to be the only one who understands you.
Neither, and no companion will pretend otherwise. Meera talks about habits and food rather than medicine, and anything that looks like a crisis is handed to real helplines rather than handled in character.
Meny is in development and not yet on the Play Store. This page and the Terms and Privacy Policy describe the app as it is being built.