Inspiration It always bothered us to open the fridge and throw out food we had genuinely meant to eat. That small, recurring guilt is happening in millions of kitchens every day, and it adds up to 40% of the food supply in the United States going to waste. We wanted to build the smallest possible intervention that could actually change that behavior, starting from something everyone already has: a grocery receipt.
What it does Parsley turns your grocery receipt into a live pantry. Snap a photo of your receipt and the app reads it, identifies what you bought, estimates how long each item stays fresh using USDA data, and suggests recipes based on what's closest to spoiling. Instead of guessing what's in the fridge or letting produce wilt in the back, you cook what you have, in the order it needs to be cooked.
How we built it We brainstormed the concept in Cursor, mocked up the UI in Google Stitch, and then synthesized everything using Claude Code and Claude Opus to generate and iterate on the actual implementation. GitHub kept the team in sync throughout. The final product is a single index.html with app.js and styles.css, deliberately lightweight, with AI API calls handling the receipt parsing and recipe generation, and local storage persisting the user's pantry between sessions.
Challenges we ran into Three big ones:
API access. Getting our API key wired up correctly took longer than expected and blocked progress until we fixed it. Receipt recognition. We started with traditional OCR (optical character recognition) but it struggled with real-world receipt photos like crumpled paper, faded ink, and weird fonts. We pivoted to using an AI vision API to read the receipt image directly into text, which was far more reliable. Food expiration accuracy. Our first version defaulted every item to a flat five-day expiration, which was useless. We rebuilt that layer to look up real shelf-life data from the USDA database so the "cook this first" recommendations actually meant something.
Accomplishments we're proud of We built an end-to-end working product in one hackathon. A user can take a picture of a receipt, upload it to the site, and with a little background magic get back an itemized pantry, realistic expiration windows, and recipe suggestions ordered by urgency. The full loop works. That's the hardest thing to pull off in a hackathon, and we did it.
What we learned Simple is better. We almost over-engineered with heavy OCR libraries and a complex backend, but a single HTML file with AI API calls and local storage turned out to be more reliable, faster to iterate on, and dramatically more demo-friendly than anything heavier would have been. The lesson: when you have limited time, let the AI do the hard parts and keep your own code minimal.
What's next for Parsley A mobile app. The whole product is about snapping a photo of a receipt, which you're almost always doing from your phone, not a laptop. Moving Parsley to native iOS and Android is the obvious next step so the experience matches the moment of use: you're standing at the checkout, you snap the receipt, and your pantry updates before you've even left the store.
Log in or sign up for Devpost to join the conversation.