For technical users
Bank Statement Parser
A parser, not just a converter: statements go in, structured transaction records come out — normalized dates, signed amounts, reconstructed balances and per-row confidence — through the browser or the JSON API.
Drop your bank statement here
PDF, JPG or PNG · up to 25 MB
No signup required · Files deleted after processing · How files are handled
Works with statements from 1,000+ banks
Parsing pipeline
Parsing runs as a staged pipeline rather than a single pass, which is what makes unfamiliar layouts tractable.
- Text extraction — embedded PDF text, with OCR fallback
- Layout detection — identifies the bank and statement variant
- Row parsing — bank-specific rules first, generic table reconstruction as fallback
- Normalization — dates to ISO 8601, amounts to signed decimals, currency detection
- Validation — running balance cross-check and confidence scoring
Output schema
Every parsed transaction carries date, description, debit, credit, signed amount, balance, currency and confidence. The same schema is returned by the API, exported to CSV/XLSX, and shown in the preview table — one contract everywhere.
Frequently asked questions
How does the parser handle banks it hasn't seen before?
A generic layout-reconstruction parser handles unknown formats by inferring column boundaries and row grouping, with balance validation flagging anything questionable.
Can I get parsed output as JSON?
Yes — download JSON from the results screen or integrate the REST API for automation.