Six faiths · Ten languages · One companion
ZindSa
English
Data, explained

Your kundli on your phone, without giving away your data

Birth date, birth time, birth place. Three of the most identifying facts about a person, handed to an app. Here is exactly what ZindSa does with them, and what it does not claim.

The sentence that would be easy to write, and is false

It would be very easy to open this page with "nothing ever leaves your phone." It reads well, it sells, and it is not true of ZindSa, or, honestly, of almost any app that has a login and a paid tier.

So that sentence does not appear anywhere on this site. What is true is narrower and worth stating precisely: your birth details do not leave your phone. The rest of this page is what that means, and where the exceptions are.

The kundli itself really is computed on your device

Your birth date, time and place stay on the phone, and the chart is calculated there. There is no kundli API in ZindSa to call, the engine is plain Dart running locally, and the founder rule it was written under is recorded in the file itself: no third-party astrology API, on-device only.

What it uses: planetary positions from JPL's "Approximate Positions of the Planets" (Standish, public domain, valid 1800–2050), the lagna from a sidereal-time formula, Rahu and Ketu as the mean lunar node, the Vimshottari dasha from the Moon's nakshatra, everything reckoned nirayana with the Lahiri ayanamsa. The wider calculation core has been tested against Swiss Ephemeris, with a maximum residual of 0.6″ measured on the slowest-moving body. Family details, floor plans and daan records are also on-device.

Not a promise, a check that runs on every build

A privacy claim that rests on a developer's good intentions is worth very little, because the code can drift away from it in a single commit that nobody notices. So this one is a build gate.

A check reads the whole calculation layer, the astronomy and calculation directories, thirty source files, as text, and fails the build if anything in there reads the system clock, the device locale, or a timezone it was not handed as a parameter. The expected value is zero and it comes from the project's own requirement register, not from the code being tested. It has been deliberately broken to confirm it goes red: two mutations were inserted, both caught, both reverted. That is what makes "computed on your device from the inputs you gave it" a statement about the build rather than about someone's intentions.

What does go to a server

Not nothing. The rule the storage layer is written under is that the server gets your name, email and phone, for login, and everything else, your dharma, your trial state, family entries and settings, stays on the device. If you buy Pro, there is a billing record too.

That is the boundary. It is small, but it is not zero, and an app that told you it was zero would be lying to you about the login screen you just used.

The two features that contact the outside world

Two things in ZindSa deliberately reach the network, and each says so on its own screen before you use it, in the app's own words.

The AI Salahkar, verbatim: “Note: to prepare the answer, your question and religion are sent to the AI service (Anthropic/Google). Do not write your name, birth details or anything private you would not share.”

The live-darshan card, verbatim: “Note: tapping play connects this phone to YouTube (Google), which then receives your IP address and what you are watching. Until you tap, nothing about this leaves your phone.”

Neither of those disclosures is buried in a policy page. They are on the screen, next to the thing they describe. And note what the first one does not send: not your name, not your birth details, not your chart. The question and the dharma, and nothing else.

And where the app admits it is not sure

Privacy is one kind of honesty; sourcing is the other, and they belong on the same page. The kundli screen carries an example worth pointing at. The naam-akshar table, the syllable traditionally associated with your Moon nakshatra, is used everywhere in practical jyotish, but its source verse could not be verified for this project: every source found was a website or a scanned image.

The app records its own status for that table as plausible and widely used, but not source-verified, and shows it to you as a suggestion rather than a rule. The instruction written beside it is not to say "the shastra says so" until a text confirms it. A privacy page that overclaims and a source note that overclaims are the same failure, and this app is trying not to make either.

Related: where your kundli data is stored in the FAQ, and how prayer times are calculated. All posts.