Hano — Family Budget Assistant
An AI-powered mobile app that logs spending from plain text and pools a whole family’s budget in one place. Live on the App Store and Google Play.
- Project type
- Mobile App
- Platform
- App Store · Google Play
- Languages
- 7
- Plans
- Free · Pro · Max
FIG. 01Screens
440 × 956
01
Hano overview screen — the household's net position, savings rate and latest transactions
02
Hano finance assistant — logging an expense in plain text and answering "what did we spend this month?"
03
Hano receipt scan — the merchant and total read off a receipt, held for confirmation
04
Hano month-end report — net summary, the assistant's commentary and category breakdown
05
Hano transactions list — instalment counters, upcoming payments and member filters
- Problem
- Budget apps make you fill out forms; most people quit after a few weeks.
- Solution
- Type "250 groceries" and the assistant does the rest — logging now takes seconds.
The real problem with budgeting apps isn't math, it's friction. Amount, category, date, account — filling a form for every purchase is why people stop opening the app after a couple of weeks. And a household budget isn't a solo job: partners spend separately, and nobody sees the total.
Solution: I built Hano around a single rule — let people type the way they think, and let the app figure it out. Type "250 groceries" or "salary landed 65,000" and a Gemini-based assistant extracts the amount, category and direction, then creates the entry. Questions like "what did we spend this month?" are answered from the same input box.
When typing is still too much, the same box opens two more paths: a sentence spoken into the microphone is transcribed, and a photographed receipt gives up its merchant, total, currency and instalment count. Both land on a confirmation step before anything is written, and neither the photo nor the audio is kept after processing.
The second layer is the household pool: users invite their partner and family, everyone logs their own spending, and the totals merge into one screen.
Architecture:
- Mobile: React Native + Expo (expo-router), TypeScript, Zustand + TanStack Query, Skia-based charts
- Backend: Supabase — PostgreSQL with row-level security and Deno Edge Functions (assistant, notifications, household invites, subscription webhooks)
- AI: An assistant layer on Gemini that turns natural language and receipt images into structured transactions
- Entry paths: The in-app box, voice dictation, receipt scanning, Siri and Shortcuts on iOS, icon shortcuts on Android, and Telegram
- Subscriptions: Free / Pro / Max tiers with store-side validation through RevenueCat
- Security: Client-side end-to-end encryption for sensitive fields (tweetnacl + scrypt); keys never leave the device
- Reach: 7 languages, light/dark themes, multi-currency with automatic conversion
The hard parts: With natural-language logging, the interesting problem isn't calling the model — it's what happens when it misreads you. Assistant output is validated against a schema, and low-confidence results surface a confirmation step; asking for one extra tap beats silently writing the wrong number. Receipts sharpen this: subtotal, tax and discount lines must never become transactions of their own, and the only number read is the grand total. End-to-end encryption had to be designed alongside multi-device sync and household sharing, and key distribution ended up being the most time-consuming piece.
Outcome: The app is live on the App Store and Google Play. Product, infrastructure, store submission and legal pages were all built and shipped by me.