Lekhit
Your UPI history, sorted automatically. A private personal finance and budgeting app built for UPI, with splits built in.
- Role
- Solo — product, design, engineering
- Timeline
- Jun 2025 — Present
- Platforms
- Android, Web
- Context
- Independent product
- Expo
- React Native
- Tamagui
- Reanimated
- Gesture Handler
- Supabase
- Postgres RLS
- Turborepo
- Bun
- Next.js
- Cloudflare Pages
01
Problem
Most of India pays with UPI now, and almost nobody knows where their month went. The data exists, scattered across PhonePe, Google Pay and bank statements, but turning it into a budget means typing every payment by hand. So people stop after a week.
Splitting is a second, separate problem. The Goa trip lives in one app and your own spending in another, and the two never agree. Lekhit puts both in one place and does the sorting for you.
02
Constraints
- Team
- Just me, so every system has to be cheap to run and easy to come back to after a month away.
- Trust
- It's people's money. The less of it I hold on a server, the less there is to leak.
- Network
- Patchy mobile data is normal. Logging a payment in the metro has to work every time.
- Habit
- If logging takes more than a few seconds, people stop. Typing has to be the exception.
03
Key decisions
Your ledger lives on your phone
- Chose
- On-device personal ledger
- Over
- Syncing everything to a server
Personal transactions are stored on the device. The app opens straight into your balance with no login wall and no spinner, works underground, and I can't read anyone's spending, by design rather than by promise.
- What it cost
- Backup is in the user's hands for now: CSV export today, cloud backup later.
- What it bought
- Privacy that's structural, instant loads and full offline capture.
Get data in without typing
- Chose
- Statement import + share-to-log
- Over
- Reading SMS or connecting to banks
Two ways in. Import a PhonePe, Google Pay or bank statement (including password-protected PDFs), and every payment lands categorised, with duplicates skipped and self-transfers not counted as spending. Or, right after paying, share the UPI receipt to Lekhit and amount, merchant and time are filled in. Fix a merchant once and it's fixed everywhere.
- What it cost
- It isn't fully real-time: you import or share, Lekhit doesn't watch.
- What it bought
- No SMS permission, no bank credentials, and a habit that takes seconds.
Only shared data leaves the phone
- Chose
- Supabase for splits, with Row Level Security
- Over
- Peer-to-peer sync, or no shared state
Splits need everyone to see the same numbers, so they're the one thing stored with your account. Row Level Security ties every row to group membership in the database, so a bug in the app can't expose another group's balances. Sign-in is Google or Apple: no passwords to store.
- What it cost
- Two storage models in one app, and RLS policies that are harder to test than app code.
- What it bought
- Splits that follow you to a new phone, with access enforced where it can't be bypassed.
Colour has meaning, so it gets rules
- Chose
- Tamagui with typed semantic tokens
- Over
- A free palette per screen
In a money app, green means money in and red means money out. If the brand colour were green, every button would read like a deposit. So green and red are reserved for money, and actions get their own colour that never collides with them. Light and dark themes are derived from the same tokens.
- What it cost
- Less freedom when designing a one-off screen.
- What it bought
- Nobody, including future me, can ship a green 'Delete' button. The types won't allow it.
A bottom sheet that knows what you meant
- Chose
- Custom sheet on Reanimated + Gesture Handler
- Over
- An off-the-shelf sheet library
The sheet has to scroll its own content and be dragged down to dismiss. Library sheets guessed wrong about which gesture you meant. Mine hands the pan to the sheet only when the inner list is scrolled to the top, so the two never fight.
- What it cost
- I own gesture edge cases forever.
- What it bought
- The most-used surface in the app feels native.
The habit is never paywalled
- Chose
- Free with no transaction caps, plus a one-time Lifetime plan
- Over
- A capped free tier or a subscription
Some apps cap free users at a handful of transactions a day. That breaks the whole point: incomplete data shows no patterns. Logging, sharing, splitting and budgeting are free with no limits. Lifetime adds unlimited statement imports and deeper insights, paid once.
- What it cost
- Less recurring revenue than a subscription.
- What it bought
- A free tier people can actually live in, and a paid plan that's an easy yes.
One repo, every surface
- Chose
- Turborepo + Bun monorepo
- Over
- Separate repos per platform
The Expo app, the Next.js site on Cloudflare Pages and a shared design-system package live together. Types, tokens and validation are written once.
- What it cost
- More up-front tooling: workspaces, build caching, Metro resolution quirks.
- What it bought
- One change to a token or schema reaches every surface in the same commit.
04
Outcome
The Android beta is open at lekhit.com.
- SMS or bank permissions needed
- 0
- To log a payment from a UPI app
- 1 tap
- Free transactions
- ∞
05
What I'd do differently
I'd design the backup and restore story before launch, not after. Keeping the ledger on the device was the right call, but "export a CSV" is a developer's answer to losing your phone. I'd also put the RLS policies under automated tests on day one instead of checking them by hand.









