I built a self-hosted family finance app with Claude Code in 9 days. Here's what it cost and what I learned

Description and use

Over nine days, Rodion built a self‑hosted family finance app using AI code generation, integrating multiple bank APIs, classifying spending, and tracking assets. The build cost ~$2,700 in tokens and runs for ~€10/month.

Problem

Managing household finances across multiple banks and countries was manual, delayed, and error‑prone, requiring a monthly Excel review that often ran late.

Outcome

In nine days, 643 commits, ~200 PRs, 61 migrations, and 659 tests were produced, delivering a production‑ready app that costs ~$2,700 in API tokens (72% for context) and €10/month to run.

Workflow

• Set up a GitHub repo and open‑source the project
• Integrated Monobank via personal API token and foreign banks via Enable Banking aggregator
• Implemented transaction import, filtering, and classification logic
• Built a Telegram bot for uncertain payments
• Created analytics dashboards and asset tracking views
• Configured PWA, health checks, backups to S3, and receipt handling
• Deployed to a VPS with Docker and PostgreSQL
• Iterated over 9 days using AI agents for code generation and debugging
• Monitored costs and optimized token usage
• Maintained the system for two months of daily use

Evidence

Nine calendar days, 11–19 September. In that time the repository gained 643 commits, around 200 merged pull requests, 61 database migrations and 659 tests.

Source

Had I paid for tokens at API prices, this build would have cost about $2,700, and 72% of that would not have gone to writing code but to the model re-reading context

Source

The app pulls transactions from banks in several countries, classifies spending, asks in Telegram when unsure, and tracks all assets in one currency.

Source