Why this exists

Eating vegan at restaurants that are not vegan should not feel like a research project. For most diners it does. Menus are written for the average customer, allergen sheets are buried three clicks deep on a corporate site, and the line cook on shift tonight does not always know whether the bread on the burger bun has whey in it. The result is the familiar moment where a vegan diner sits at the table, reads the menu twice, and quietly orders fries and a side salad.

Vegan Recon attacks that gap directly. Paste a restaurant menu URL and the app reads every dish on the menu, checks every ingredient against a vegan rule set, and hands back the items that are already plant-based or work with a small swap. Each item ships with the exact wording to use at the counter so the kitchen can say yes the first time. The goal is to make non-vegan restaurants navigable without having to memorize a playbook for every chain.

A scan is a four-step pipeline. First, a headless browser visits the menu URL and pulls down the rendered HTML, which handles single-page menus that lazy-load their dish list after page load. Second, a parser walks the DOM and extracts structured dish records with names, ingredients, allergen tags, and prices wherever the restaurant publishes them.

Third, a language model audits each dish against a vegan rule set built from years of menu reading - the dairy traps in sauces and breads, the swaps every chain knows how to do, the ingredients that have to be confirmed at the counter even if the menu does not list them. The audit produces a confidence score per item, which gets calibrated against ground-truth data from human curators. Fourth, the scan caches its results so the next visitor to the same restaurant gets the same answer in under a second without re-running the pipeline.

The scan is opinionated where opinions matter. We do not silently include items that depend on a fryer that the kitchen also uses for chicken; we surface that as a check to confirm at the counter so the diner gets to make the call. We treat honey as non-vegan because the dominant North American convention is to do so, but we flag honey explicitly so a vegan who eats honey can override.

Vegan Recon is built by True North Technology, a small, independent software outfit run from Michigan. There is no venture capital, no advertising-driven engagement model, and no plan to introduce a paywall around the basic scan. The free tier handles three scans a month and the entire saved catalogue of every scan anyone has ever run. The paid tier unlocks unlimited scans and a few quality-of-life features for users who use it weekly.

The reason it stays this way: the work is funded by the person who wanted the tool to exist. That covers servers, language-model fees, and the occasional design contractor who keeps the brand sharper than a solo founder can manage. Nothing about the product strategy depends on aggressive growth or harvesting user data, and that lets us optimize for the only metric that matters - did the diner get the right order at the counter.

The scan is right most of the time and wrong some of the time. We publish a confidence score on every item so the uncertainty is visible, and we maintain a check-before-you- order section per item that surfaces the questions a diner should ask the counter before ordering. That second layer is where the model leans on human knowledge instead of pretending to know.

The places the scan is least reliable: locally-owned restaurants without a structured online menu, regional chains that change their ingredient suppliers between markets, and anywhere the kitchen reformulates a dish without updating the website. When you order a Vegan Recon-flagged item and the kitchen confirms something the scan missed, please send us a correction. Every correction trains the next round of scans.

Vegan Recon is not a medical-allergen reference. We surface allergen information when the restaurant publishes it, but the canonical source for a serious allergy is always the restaurant's own allergen sheet plus a conversation with the kitchen.

Two near-term focuses: more chain coverage and better local coverage. The chain queue is curated by hand and prioritizes restaurants that vegan diners visit most, weighted by the number of scan requests we get for each. Local coverage depends on whether the restaurant publishes a structured menu online; we are working on a graceful degradation path for restaurants whose menu is a JPEG photographed at lunch service.

If there is a chain you want covered, an item we got wrong, or a feature you want to see, the email is truenorthtechnologymi@gmail.com. We read everything.