Litra
A WhatsApp-based AI financial tracker that lets you log transactions, manage budgets, and track loans through natural conversation.
Overview
Litra is a WhatsApp-based AI financial tracker built for people who want expense tracking without a separate app. You send a message, Litra understands it, and the transaction is logged. That is the whole interaction model.
The backend is written in Go, using Whatsmeow for WhatsApp integration and Gemini for natural language understanding. Transactions are stored in PostgreSQL, with subscription billing handled through Midtrans.
What I focused on
- Low friction transaction record methods with natural language parsing so users can write "bayar kos 800k", also with receipt pictures and voice notes.
- Transaction editing via WhatsApp reply, so corrections feel as easy as the original log
- Recurring transaction support for fixed monthly expenses
- Loan and piutang tracking with a clear repayment flow
- Subscription plan upgrades with proration so billing stays fair
Outcome
Litra is the product I have kept iterating on the longest. Each feature comes from a real gap in how I personally wanted to track money: no app switching, no manual categorization UI, just a message.
The system is small but the interaction surface is high. Getting the NLP layer right for Indonesian financial language was the most interesting engineering problem in the project.