I built FinKhoj because every bank-statement analysis was eating hours on clerical work — cleaning up statements, chasing the same party across different spellings, figuring out which entries were the same money counted twice. Two mistakes taught me the most: 1. I trusted the extracted data too early. Now the app double-checks every single row against the statement's own running balance before anything else happens — so a bad extraction gets caught immediately instead of quietly becoming a wrong answer. 2. I used to match transfers across accounts if the date and amount looked close. Turned out that's not real proof — those reference numbers already contain the date, so same-day transfers naturally look similar. Now a transfer only counts as linked when both statements show the exact same full reference number. About half my old matches turned out to be wrong. I'm not a developer — I built this with Claude, using what I know from doing this work by hand for years. Would love to hear: if you do this kind of work manually, where does your process usually break down.

Tarang

Trace Transactions. Connect Parties. Follow the Funds

Trace Transactions. Connect Parties. Follow the Funds