The Shoebox Accountant That Runs On Your Own Server

TaxHacker will take a shoebox of receipts and a folder of invoice PDFs and turn them into categorized, exportable transactions in a structured database, the kind of manual data entry a small business owner either grinds through themselves on a Sunday night or pays a bookkeeper twenty to forty dollars an hour to do. That is the entire pitch, and it happens to be a real one. No dashboard tour, no reimagined workflow language, just software that reads paper and PDFs and hands back clean rows you can drop into a ledger.

Here is what actually changed. vas3k/TaxHacker is an open source, self-hosted app that runs an AI analyzer over receipts, invoices, and bank transactions, sorts them against categories you define, and stores the result in a Postgres database that lives on your own server, not someone else's cloud. It has been climbing steadily in stars and commit activity this year, the kind of curve that separates a weekend side project from something people are actually installing and using. It ships under a permissive open source license, which is the detail that matters most to a business owner, since it means you can run this in production without a lawyer reading fine print first.

Why this lands for a small operator has less to do with the software and more to do with where the hours actually go. Receipt and invoice entry never gets its own line item on a budget. Nobody puts "three hours keying in vendor invoices" on a P&L, it just eats an evening, or it quietly pads a bookkeeper's monthly bill by a few hours nobody tracks closely because the alternative feels like more trouble than it is worth. Tools like Expensify or Dext solve the same problem with a per seat subscription and someone else's servers holding your financial documents. A self-hosted version flips that trade, you keep the data, but you also inherit the running of it. Either way, this is the exact rung an operator graduating off spreadsheets is standing on right now, the point where manual entry stops being a personality trait and starts being a cost center worth naming. A tool like this does not replace an accountant, and it should not try to. What it replaces is the keying in, the part where a person opens a dozen PDFs, transcribes vendor name and amount and date by hand, then guesses which bucket the expense belongs in.

Now the honest take, because the project's own readme says it plainly and so should this piece. TaxHacker is early software, use it at your own risk. Running it means standing up Docker and a Postgres instance, and for most shops in the twenty to two hundred person range, that means either finding the one person on staff who has done that before or paying someone outside to do it, which quietly changes the math on what a free tool actually costs. The commit history shows the deeper tell. The same maintainer who pitched this as a privacy first, self-hosted alternative to Expensify has a commit on record titled enable saas subscription, and the project's build activity confirms development is moving fast enough that a hosted paid tier landing on top of a free self-hosted tool is not a hypothetical, it is already in motion. That is not a scandal, maintainers have to eat, but it is precisely the pattern that has burned plenty of operators before. Adopt the free thing, get dependent on the workflow it creates, then watch the pricing model shift under your feet once the user base is big enough to monetize.

None of that means skip it. It means read the commit log of anything you self-host the same way you would read the terms of service on anything you subscribe to, because open source does not mean the business model is finished, it just means you get to watch it get built in public before it gets built around you.