Statement Autopsy

Bank statements hold the answer to 'where does my money go?', but they're a wall of cryptic UPI narrations. The tools that read them want your login or your data on their servers — a lot to ask for something this personal.

1. 01Privacy as the architecture, not a policy: parsing and every check run in the browser — no server, no database, no account, no AI service.
2. 02Read real bank exports: header detection, wrapped narrations, newest-first files, Dr/Cr columns, card statements — and the running balance double-checks every sign.
3. 03Only honest numbers: the headline adds up concrete fixes (avoidable fees, overlapping subscriptions, two home-cooked meals a week), never 'all your delivery spend'.
4. 04Every finding is traceable — one click shows the exact payments behind it, and re-labelling a merchant re-runs the whole report.

• Readers for CSV, Excel and PDF (pdf.js, rows rebuilt from text positions), including password-protected e-statements.
• Merchant recognition for ~120 Indian merchants plus UPI, NEFT, IMPS, ACH and POS narration parsing; unknown merchants still group by name.
• Six checks: subscriptions (overlaps, price rises), bank fees, double charges, food delivery, small spends and cash — each with a next step and a yearly saving.
• Report with vitals, charts, an evidence locker (search, filter, re-label), categorised CSV export and print-to-PDF, in light and dark mode.

• Next.js 16
• React 19
• TypeScript
• Tailwind CSS 4
• GSAP + Lenis
• pdf.js
• SheetJS
• Papa Parse
• Vitest
• Playwright
• Vercel