The problem

Every month you sit down to reconcile books and discover your bank exports PDFs, not CSVs.

The one-line fix

curl -X POST https://ashdocs.io/api/v1/tools/bank-statement-to-csv \
  -H 'X-API-Key: dr_test_...' \
  -F 'file=@Chase_January.pdf'

That's it. You get back a CSV with date, description, debit, credit, balance — reconciled to the opening + closing balances printed on the statement.

Why not tabula / camelot / OCR?

Because they don't understand statements — they see tables. Rows split across pages get treated as separate tables. Footnotes leak into the data. Credit and debit columns get swapped when the sign convention flips halfway through.

ASHDOCS is trained specifically on bank-statement formats — HSBC, Chase, Wells Fargo, Barclays, Monzo, HDFC, all of them.

What you get back

• transactions[] — sorted, deduped, typed (numbers not strings)
• reconciliation — passed=true only when your closing balance ties out
• confidence — per-field 0-1 score, low-confidence rows flagged for review
• files — CSV + optional QBO / OFX exports

Next step

Grab a free sandbox key and try it against your last statement. If it doesn't reconcile, tell us — we'll fix it.