Personal Finance Stack

What's in this pack

This pack collects the six self-hostable finance applications that, together, replace YNAB, Mint, QuickBooks Self-Employed, Stripe Billing, and your shoebox of receipts. Every entry is open source and runs from a Docker image. The AI receipt parser is the glue: it takes a phone-camera photo and emits a structured transaction that the other tools accept.

Firefly III is the load-bearing entry: it runs double-entry accounting, supports multi-currency, and accepts CSV imports from every major bank. The other five plug into it.

Why own your finance tooling

Money apps are the longest-lived software you'll keep. A receipt from 2018 might matter in a 2026 audit; a SaaS app you used in 2018 might be dead, acquired, or paywalled today. Self-hosting your money stack is about three things:

• Long retention without vendor risk. Your transactions live in Postgres on your disk. Ten years from now they will still open.
• Privacy. Mint and Personal Capital sell aggregate data to advertisers. Your bank already knows what you bought; the analytics vendor doesn't have to.
• Custom rules. Firefly III's rules engine is more flexible than Mint's category mapper. You can categorize a transaction by merchant + amount + date pattern, not just by a single string match.

The cost: you operate a small Postgres + Redis + Docker stack and back it up. Firefly III alone has 5 years of stable releases, so the operational burden after first install is low.

Install in one command

# Install the entire pack
tokrepo install pack/personal-finance-stack

# Or pick the apps you need
tokrepo install firefly-iii
tokrepo install actual-budget
tokrepo install ai-receipt-parser

The TokRepo manifest provides a single Docker Compose with shared Postgres, the AI receipt parser pre-wired to Firefly III's import API, and a backup script that runs pg_dump nightly to a local rclone-friendly directory.

Common pitfalls

• Bank CSV format drift. Banks change their CSV column order without notice. Pin your Firefly III importer profile to a specific format and re-test after every quarterly statement download.
• Multi-currency ambiguity. Firefly III handles multi-currency well, but you must set the base currency on first install and cannot change it later without a manual data migration. Pick the currency you'll file taxes in.
• AI receipt parser hallucination. Even a tuned model occasionally invents a vendor name from a blurry receipt. Always route parser output through a human review queue before posting to Firefly III. Don't auto-commit.
• Lago for personal use is overkill. Lago is built for usage-based SaaS billing. If you don't sell metered software, skip it. We left it in the pack because freelancers selling per-hour or per-API-call services need it.
• Invoice Ninja's tax engine needs configuration per region. Default install assumes a US sales-tax model. EU VAT and India GST are supported but require explicit setup before sending the first invoice.

Common misconceptions

People often think a self-hosted finance stack means giving up bank syncing. It doesn't. Firefly III imports CSV from all major banks. Power users add nordigen / GoCardless Bank Account Data, Plaid via firefly-iii-data-importer, or scheduled scrapes — depending on country. The model is closer to "every other day, batch import" than "real-time as in Mint." For most households that's a feature, not a bug: you actually look at your transactions.

Another misconception is that you need accounting expertise to run Firefly III. You don't. The double-entry model surfaces only when you choose to look. Day-to-day usage looks like Mint: a transaction list, a category picker, a few dashboards. The double-entry rigor matters when you go to file taxes, transfer between accounts, or reconcile a credit-card statement — those flows that single-entry apps quietly get wrong. Treat the first two weeks as bank import + categorization muscle memory; the accounting depth waits for when you need it.