Skip to content
DocsGuides

How parsing works

There is no supported bank list, no per-issuer template and nothing to configure before you start.

Supported banks#

All of them, as far as the API is concerned. Nothing is set up per institution, so there is no list to check your customers against.

  • A statement from an issuer we have never seen reads like one we test against daily.
  • A bank redesigning its statement does not break your integration.
  • There is no queue to join and no per-institution work to schedule.
  • UK, US, Canadian, Australian, Indian and European statements are all in regular use, in any currency.
  • Business accounts, joint accounts and statements from closed accounts work the same way.

What to send#

  • A statement. A payslip, invoice, receipt or loan history returns 422 not_a_bank_statement, free.
  • In any form the customer has it: the PDF the bank issued, a scan, a phone photo, or a CSV, OFX or QIF export. See Files you can send.
  • Up to 50MB. Up to 60 pages in one request, or up to 500 with async.
  • Password-protected is fine. Send the password in a multipart part named password.

Long and multi-account statements#

A file holding six months of rows, or four accounts, is one document. Nothing is truncated to fit. If part of a long statement could not be read, verification says so. See Accounts and periods.

Accuracy#

Before a document is returned, the extraction is checked against the statement. Where the issuer prints an opening and a closing balance, the transactions have to account for the movement between them exactly. See Verification.

Tampering#

Every document also says whether the file was edited after the bank made it, modified after it was created, or sent to you before. See Authenticity signals.

Something wrong or missing on this page? [email protected]