Upload

Each file is extracted, matched against the PO table, and stored.

01Invoice files

Drop invoices here

PDFs, images (PNG/JPG), or text/JSON mocks

02Sample files

Six pre-built invoices covering every scenario. Download one or several, then drop them into the upload area above.

Acme Office Supplies

→ MATCHED

JSON mock · matches PO-1001 exactly

01-acme-matched.json

Globex Industries

→ AMOUNT MISMATCH

Text invoice · €9,408 vs €8,400 expected (+12%)

02-globex-amount-mismatch.txt

Initrode Software

→ SUPPLIER MISMATCH

JSON mock · PO-1003 is for 'Initech', not 'Initrode'

03-initrode-supplier-mismatch.json

Stark Logistics

→ PO NOT FOUND

Text invoice · references PO-9999 (doesn't exist)

04-stark-po-not-found.txt

Acme Office (duplicate)

→ DUPLICATE

Same invoice number as #1 — upload after #1 to trigger

05-acme-duplicate.json

Wayne Enterprises

→ MATCHED

Real PDF · tests Claude vision extraction · matches PO-1005

06-wayne-real-pdf.pdf

How extraction works

PDFs and images are sent to Claude (Anthropic) for vision-based field extraction. Plain text and JSON inputs are parsed directly. The extractor returns structured fields — supplier, invoice number, PO number, amount, currency, dates — which are then matched against your purchase orders. No SAP integration needed.