The self-feeding finance tracker that does the work for you
A developer tired of manually feeding his expense tracker decided to build one that feeds itself—and in the process, saved countless hours. The result is FinMan, a free personal finance application that now imports bank statements with AI, reads utility meters and fuel receipts from photos sent to a bot, and even calculates the real cost per kilometer driven. What began as a simple Perl script on the backend has evolved into a system designed to eliminate the manual labor that frustrates users of traditional finance apps.
From tedious typing to effortless tracking
Popular expense trackers often require users to enter every transaction by hand, categorize expenses for hours, and still miss many online card payments—the very transactions that account for most spending. The developer behind FinMan found that balancing feeding the tracker with actually understanding his finances became unsustainable. His solution: an app that works the way he lives. The first major improvement came from integrating Monobank’s webhook API, which automatically records card transactions the moment they occur—no typing, no forgetting. This set the foundation: if a step requires discipline, it will eventually fail. Automation had to be the core principle.
AI that handles any bank statement—no new code required
When adding support for other banks, each with its own statement format—CSV, PDF, or custom layouts—the developer initially wrote custom parsers for Privatbank, PUMB, Raiffeisen, Revolut, Wise, Erste, and N26. But maintaining separate parsers was unsustainable. The breakthrough was a declarative extraction engine powered by AI, which recognizes the structure of a statement it has never seen before. Now FinMan imports any bank statement without new code. The real challenges turned out to be the “boring” parts: ensuring stable payment IDs to prevent duplicate entries and handling currency conversions per transaction row.
Photos as the ultimate input: meters, receipts, and odometers
Two monthly rituals consumed evenings: reading utility meters and calculating car costs per kilometer. Now, users can simply send a photo of a meter or fuel receipt to a Telegram or WhatsApp bot. AI reads the meter digits, identifies the utility service, computes the difference, and generates the bill. For car expenses, a photo of the fuel receipt and odometer delivers accurate fuel consumption and cost per kilometer. The system uses a two-phase recognition pipeline: first, a universal prompt classifies the photo type (meter, receipt, odometer), then a specialized prompt processes the details. Both Telegram and WhatsApp bots share identical logic, with routing handled at the API level—making future platforms easy to add.
Source: DEV Community. AI-assisted editorial synthesis — TechnoExpress.