AI-Powered Receipt Splitting SaaS with Stripe Integration

This product turns the most annoying part of eating out into three taps: photograph the receipt, tap who had what, send everyone their number.

The user photographs a restaurant bill. Claude vision reads every line into structured, validated JSON. The user corrects anything the AI got wrong (every field is editable), adds the people at the table, and assigns each item to one person, several, or everyone. The app then works out exactly what each person owes, sharing tax, service, tip and discount proportionally, and produces a plain-text summary ready for WhatsApp.

It is a working product, not a prototype: email and password accounts, receipt history, a free plan with a monthly scan quota, and Premium subscriptions wired through Stripe Checkout, Customer Portal and signature-verified webhooks. Money is calculated with decimal arithmetic and largest-remainder rounding, so per-person totals always add up to the bill exactly, to the cent.

The backend is hardened well beyond a typical side project: PostgreSQL Row Level Security on every table with a least-privilege runtime role, nonce-based CSP, HSTS, CSRF origin checks, rate limiting, streaming upload limits and magic-byte image validation. Receipt photos are processed in memory and never stored. 92 unit tests and 48 end-to-end tests cover security, payments, accessibility and responsive layout, and run in CI.

It is pre-revenue: deployed, tested with real receipts, and never marketed. You are buying a finished codebase and a live deployment, not a customer base.