Invoice to JSON

Extract invoice number, dates, parties, line items and totals from a PDF invoice into structured JSON - with arithmetic validation and low-confidence fields flagged for review.

Multi-file input
Long-running / async
Invoice field-extraction workspace

This tool detects standard invoice fields - invoice number, dates, vendor and bill-to, totals, payment terms - by matching common multi-language label wording and capturing the nearest value, detects the line-item table the same way, and validates that quantities × unit prices and subtotal + tax reconcile to the stated total. It does not run OCR - a scanned page with no text layer is flagged and skipped rather than guessed at. It only reads the file; nothing is changed.

Simple workflow

How it works

1
Add invoices

Add one or more PDF invoices to the list.
2
Extract

Extract a file and review its fields, verifying low-confidence ones against the original page.
3
Download

Download the structured JSON.

Frequently asked questions

Does this remember a vendor's layout after I correct a field?

Not yet. Every extraction is independent - a correction only fixes that file's export, it doesn't change how future invoices from the same vendor are read. Persistent per-vendor templates may come later as a separately-scoped feature.

Does this OCR scanned invoices?

No. A page with no extractable text layer is flagged and skipped, not guessed at through OCR.

What happens when the arithmetic doesn't reconcile?

The mismatch is flagged for review - the extracted numbers are never silently corrected.

Stay ahead of your document work

New tools and thoughtful product updates, delivered occasionally. No noise, no spam.