What this tool does
Invoice data extraction reads the key fields out of a batch of PDF invoices at once — invoice code, number, issue date, amount, tax and total — and turns them into a single list you can hand to finance, exported as CSV or Excel.
It also takes care of the most common reimbursement mistake: duplicates. The same invoice downloaded twice and claimed twice is a real problem, so this tool flags repeats by invoice code + number and can drop them for you.
Everything runs in your browser; invoices are never uploaded. Invoices are sensitive financial data, so this is not just a slogan — open your devtools network panel and confirm there are no upload requests.
How to use
- Drop your invoice PDFs Multi-select and drag-and-drop are supported, up to 50MB per file — dozens at once is fine. Order does not affect the result.
- Check the summary table You get a table first: one row per invoice, totals at the bottom, duplicates highlighted. Scan for blank cells — a blank means that field could not be read.
- Export CSV or Excel The CSV carries a BOM so Excel opens Chinese headers correctly; the Excel file stores amounts as real numbers, ready for SUM and pivot tables.
FAQ
Are my invoices uploaded?
No. Parsing and summing happen entirely in your browser; the network panel shows no upload requests. Invoices are sensitive, so this tool never talks to a server by design.
Why is a field blank for some invoices?
Usually that invoice has an unusual text-layer layout, or no text layer at all (a scan or photo). This tool would rather leave a cell empty than guess. Fill it in by hand, or re-download the PDF from the issuing platform.
Which invoices are supported?
PDF e-invoices with a text layer: VAT special and general invoices, vehicle sales invoices, and fully-digital invoices in PDF form. Scans and photos of paper invoices are not supported, nor is OFD.
How are duplicates detected?
By invoice code + number: when both match, it is the same invoice. Fully-digital invoices have no code, so the number alone is used. Invoices whose number could not be read are excluded from dedupe to avoid false matches.
Can it read the buyer and seller names?
No. "Name" appears twice on an invoice (buyer and seller) and the text layer rarely disambiguates them reliably, so this tool only outputs fields it can locate with confidence.
How many invoices can I process?
No fixed count limit, but each file must be under 50MB and the total is bounded by browser memory. A few hundred is usually fine; split larger batches.
Limits
PDF with a text layer only — scans, photos and OFD files cannot be read.
Buyer/seller names and line items are not extracted, and there is no tax-authority verification (a local tool cannot call the tax service).
Original invoices are never modified; only a summary list is produced.