FairSlice
An app for splitting shared costs such as restaurant bills, rent, trips and game nights. It reads receipts, handles over 150 currencies and turns a tangle of debts into a short list of payments to settle up.
- Type
- Our own product
- Year
- 2025
- Status
- Live at fairslice.app
The problem
Splitting costs with friends, roommates and groups was still awkward in 2025. We wanted one app that covered itemized restaurant bills, poker-night settlements, recurring rent and travel in several currencies.
Venmo and PayPal recorded payments, but they did not keep track of who owed what across several expenses. Group chats turned into “did you pay me?” arguments that went nowhere.
What we built
An app for splitting shared costs, with four ways to split, AI-assisted receipt scanning, game-night settlements, a short list of payments to settle a group, scheduled reminders and group balances that update live.
Splitting a cost
A cost can be split equally, with rounding so nobody is short a penny; by exact amounts, when one person had the expensive dish; by percentage, for an ownership stake or agreed ratio; or by shares, such as two against one. Several people can pay toward one expense (Alice $60 and Bob $40), a restaurant bill can be split item by item, and a recurring setup can be saved as a template.
Receipts and categories
From a photo of a receipt, Anthropic’s Claude AI reads the merchant, each item’s quantity and price, the tax, tip and total, and the currency, and scores how confident it is. You assign items to people, and the tax and tip split in proportion, with nothing typed in.
Expenses are sorted into 16 categories, such as food, transport and housing. Keyword matching sorts most of them instantly, and Claude Sonnet decides when a description is ambiguous.
Game nights
Templates cover eight games: Texas Hold’em, rummy, teen patti, spades, blackjack, baccarat, mahjong and 21. FairSlice tracks each player’s buy-in and rebuys, scores in chips (poker) or points (rummy and mahjong), and works out the settling payments in one tap.
Settling up
If Alice owes Bob $20, Bob owes Charlie $15 and Charlie owes Alice $10, FairSlice reduces that to a short list of transfers, never more than one fewer than the number of people. It pairs the largest debtor with the largest creditor, repeatedly.
A payment counts only when the receiver confirms it, so both people agree on what was paid. People can settle by cash, Venmo, PayPal, Zelle, UPI, Apple Pay, Google Pay, Interac, Revolut, Wise or crypto.
Reminders
Reminders go out automatically, more often as a debt gets older.
| Age of the debt | Reminder |
|---|---|
| Under 7 days | None |
| 7 to 30 days | Only if the amount is $50 or more |
| 30 to 90 days | Every 14 days |
| 90 to 180 days | Every 7 days |
| Over 180 days | Every 3 days |
Insights
A fairness score from 0 to 100 shows how evenly a group’s expenses are shared, using the Gini coefficient, a standard measure of inequality. FairSlice also shows each person’s spending by category, keeps snapshots of balances to show how debts age, and flags the debts that matter most by combining amount and age.
Other features
- Conversion between over 150 currencies at live rates
- Expenses between two friends, outside any group
- Weekly and monthly expenses that recur automatically
- CSV import and export, for moving in bulk and for GDPR-compliant data export
- Search across expenses, groups and friends
Result
FairSlice is live at fairslice.app. It handles over 150 currencies and the everyday cases: rent, restaurant bills, travel and recurring game nights.
Technical notes
How it is built
Written in TypeScript in strict mode, with over 250 type definitions, on Next.js 16 and its App Router, React 19 and Tailwind CSS v4. The server has 107 API routes (the addresses the app calls for data), and Zod 4 validates every input they accept. Supabase provides the PostgreSQL database, logins, and Realtime channels that push changes to a group instantly. Webhooks notify other software of nine kinds of event. Tests run on Vitest. The app is deployed on Vercel’s edge network, with daily scheduled (cron) jobs.
Security
Rate limits apply in four tiers: standard, heavy AI operations, login and data export. The server checks each request’s Origin header so another website cannot act for a signed-in user (CSRF protection).
Webhooks are signed with HMAC-SHA256 so the receiver can verify them, retried with exponential backoff (a longer wait before each attempt), and guarded against SSRF, so they cannot be aimed at the server’s own network.
Users choose whether their email address and phone number are visible, and can delete their account completely.
Next case study
Start a project
Tell us what you need built.
Send a few lines about your business and what you have in mind. We reply within one business day, book a 30‑minute call if it makes sense, and send a fixed quote in writing after that.
Or call 647-252-5352, Monday to Friday, 9 am to 6 pm Toronto time.
