Read any food label in seconds, and know if it's worth buying.

This started as a personal frustration, not a market study. I watch people across India pick a packet off the shelf and drop it in the cart without a second glance, with no idea what is inside, whether the additives are safe, or whether it is something other countries stopped allowing years ago. It is not carelessness. The label is right there in their hand, and it is effectively unreadable: microscopic type, cryptic E-numbers, and a nutrition table almost nobody is trained to read.

OpenLabel exists to close that gap. Not to lecture and not to scare, just to put a clear, honest answer in someone's hand in the two seconds they are deciding whether to buy. Knowledge first. What they do with it stays their call.

Every scanner app I tried fell down in the exact moment that mattered, standing in a shop with a packet in hand. They lean on big online databases with thin coverage of Indian products, and they need a connection that basement supermarkets simply do not have. So you scan a local biscuit and get “not available,” or a score with no reason you can check. Meanwhile a colour or preservative that is freely used here may be banned or warning-labelled abroad, and there is no way to know that from the aisle.

The product is deliberately one surface: the camera. You either scan the barcode, which is fast when the product is known, or you point it at the ingredients label and let the app read the text itself. The label path is the important one. It works on any product on the shelf, even one no database has ever heard of, and it works with the phone in aeroplane mode. A morphing reticle and a single switch move you between the two without ever leaving the screen.

The quiet engineering decision that makes OpenLabel usable is the lookup chain behind a barcode. Instead of one database and a shrug, a scan flows through four stages, each one a wider net than the last. By the time it reaches the label scanner, the question is not whether it can answer, only how.

A score people are meant to trust cannot be something they have to take on faith. So the scoring is deterministic and rule-based, never AI. It combines two independent things: the nutrition, scored with the EU-validated Nutri-Score model, and the additives of concern, each one deducting points that are capped so a single flag cannot tank an otherwise decent product. Every point up or down names the exact ingredient responsible, and the same pack always returns the same number.

The most interesting problem was one that first looked like a bug. Some clearly sensible foods were scoring badly, and the reason was that nutrition was being judged per 100 grams, flat. A pickle or chutney is brutal per 100 grams, but you eat about 15 of them. Scoring it as if you ate a full 100 was punishing food nobody eats that way.

The fix was to score nutrients per a realistic serving, using standardised reference amounts by food group, and to do it only for concentrated, small-serving savoury foods. Sweets, drinks and snacks are deliberately left untouched, so the adjustment can never quietly make junk look better. I validated it against a basket of real products before it shipped. The right answer was not a new screen. It was a more honest rule underneath.

This is the part no other app gives you in the aisle. OpenLabel carries a curated registry of additives with their status in other regions, the US, EU, UK, Canada and Australia, recorded as banned, restricted, or warning-labelled, and always naming the specific country. A whitener that is banned in the EU or a colour that needs a warning label there is not something a shopper in India can otherwise know while holding the box.

Health apps are usually where your data goes to be sold. OpenLabel is built the other way. The label photo is read on the device and never uploaded. The only thing that ever leaves the phone is a barcode number, sent to look up a product. History lives in a local database, and there is no account, no login, no analytics, no ads and no trackers. Its Google Play data-safety declaration is, literally, “No data collected.”

OpenLabel is not trying to replace your diet tracker or the internet. It fills the one gap they all leave open: a trustworthy read on a real product, in the aisle, offline.

OpenLabel is the project where I own every layer: the idea, the brand, the interface, the scoring logic, and the app itself. I designed it and built it with Claude as my engineering partner, which is exactly the kind of end-to-end product work I want to be doing more of. The judgement calls that mattered most, keeping the score rule-based, scoring by a real serving, sending nothing off the phone, were product decisions long before they were code.

It is honest about where it is. Barcode and label are the strong paths; the OCR is still being tuned against messier real-world packaging, and not every Indian product is in the offline set yet, which is exactly what the live lookup and the label scanner are there to cover.