How the invoice parser works
Select one or more invoices or receipts, as PDFs or as photos from your phone, and start the batch. Each document is read on its own: text-layer PDFs directly, scans and photos through OCR, and a language model then fills a fixed invoice template with the values printed on the document.
Documents are processed one after another and the queue shows which are done. When the batch finishes, the parsed values appear in a summary table and can be downloaded together. Each document uses the same AI data extractor with the invoice template, so single invoices can also be processed there.
What is extracted
For every document the parser returns the document type, vendor name, address and tax or VAT number, the customer, the invoice or receipt number, issue and due dates, currency, subtotal, tax, total, payment method and every line item with description, quantity, unit price and amount.
Dates are normalised to YYYY-MM-DD and amounts to plain numbers, so the result sorts and sums correctly in a spreadsheet. Fields the document does not show stay empty rather than being guessed.
Export to Excel, CSV and JSON
The Excel file has two sheets: one row per invoice, and one row per line item with the invoice number next to it. The same two tables are available as separate CSV files, saved as UTF-8 with a byte order mark so accented vendor names open correctly in Excel. JSON keeps the full nested result of every document.
The files are built in your browser from the results of the batch, so re-downloading in another format does not process the documents again. To combine scanned receipts into one PDF before archiving them, use Merge PDF.
Using it from an ERP or script
The parser is available through the same REST API as every other tool: send each document to POST /api/v1/jobs with operation extract-structured-data and options {"template": "invoice"}, poll the job and download the JSON result. The response has the same fields as the JSON export on this page.
An API key raises the file size, page and daily limits; see the API documentation for authentication and request examples.