All work
Mobile App 2026 · Appcademy

Belanja

AI receipt scanning with a credit wallet

The problem

Expense tracking fails at the same point for almost everyone: data entry. After two weeks of typing in every receipt by hand, people stop.

So the real problem was never charts or categories. It was getting the effort of recording one expense down from a minute to a few seconds.

What we observed

Before we wrote any code.

Scanning receipts with AI solves the entry problem and introduces a second one: the cost of each scan.

That is why this uses a credit wallet rather than an unlimited subscription. The user can see how many scans remain, and the system cannot run up a bill beyond what has been paid for — protection in both directions.

Every scan result stays editable before it is saved. AI reads receipts well but not perfectly, and a system that pretends otherwise will record the wrong amount with nobody noticing.

What we built

Belanja reads a shopper's receipt with an AI vision model and turns it into structured spending. Built on Fastify + PostgreSQL with a credit ledger, device-bound accounts, quota control, pluggable AI providers and an owner console for live configuration.

Before vs after

The same job, the old way and the new way.

Before

Every receipt retyped by hand
Most people stop recording after a few weeks
Unlimited AI usage makes cost unpredictable

After

Receipts are scanned; fields fill in and stay editable before saving
A credit wallet shows the remaining scans — no surprise bills
Recording one expense takes seconds instead of a minute

Results

27
API endpoints in production
12
database tables behind the wallet
Live
owner console for runtime config

Build something similar

Let's build yours.

Tell us what you're planning. Free 30-minute call, a written scope and a fixed quote — we reply within one working day.

Start a project →