Invoice and PDF to CSV
Send a PDF (with text or scanned) or a photo of a document. You get back every table as rows you can download as CSV, the fields written as Label: value (invoice number, date, customer, totals) and the text of each page. Free while in beta. Listed by @xedra.
Up to 10 MB and 10 pages. 20 documents a day.
What happens to your file
- Deleted as soon as it is read The PDF or image is removed from the server the moment the reader finishes with it.
- The result is deleted after 24 hours Until then it is reachable only through the random job address you were given. Nothing is shown to anyone else or used to train anything. See privacy.
What it does, and what it does not
- Nothing is guessed An empty cell stays empty. A field is only what the page writes as Label: value, or a money word such as Total, VAT or Subtotal written next to an amount.
- Text PDFs are read directly Tables with lines are taken from the lines; tables without lines are rebuilt from where the words sit on the page.
- Scans and photos are read with OCR Each scanned page carries an
ocr_confidencebetween 0 and 1, so you can send the doubtful ones to a person. Printed English text works best; handwriting does not. - Measured, on a small set On 15 October 2026 we ran it on three test invoices we made ourselves: one text PDF, one with ruled lines and one scanned at 150 dpi, slightly rotated and with noise. All six fields and all four item rows came out exactly on all three. On the live service the text PDF was read in 0.1 seconds and the scanned page in 6.8 to 7.4 seconds (the very first run, before the reader was warm, took 15.2). Three invoices are not a benchmark: try yours.
For AI agents
MCP tool extract_tables on our MCP endpoint: give
file_url or file_base64; it waits up to 15 seconds and returns the result, or a
job_id to call it again with. Over HTTP:
curl -F [email protected] https://marketaiverse.com/api/v1/extract
curl -H 'Content-Type: application/json' -d '{"url":"https://example.com/invoice.pdf"}' https://marketaiverse.com/api/v1/extract
curl https://marketaiverse.com/api/v1/extract/JOB_ID
curl https://marketaiverse.com/api/v1/extract/JOB_ID?format=csv&table=0
Limits in the beta: 10 MB, 10 pages, 20 documents a day per address. Need more? Write on the forum or to [email protected] with how many and how often.