- Upload
- Options
- Process
- Download
This file stays on this device.
No watermarksAI Extract
List the fields you need. We reuse named-field extract and invoice line items — not a new model. Blank or weak values are flagged for review.
How to extract a schema from a PDF
- Add the PDF. The file stays in your browser.
- Keep the defaults or type a list such as Vendor, Invoice number, Total.
- Extract. Sparse pages can be OCR’d locally.
- Check confidence, edit weak rows, download JSON, CSV, or Excel.
Features
Your schema
Type a list or add labels. We look for those words on the page.
Confidence
Empty, odd amounts, and weak dates are flagged. You can still download.
Same engine
This is Extract PDF fields plus scores — not a second extractor.
Line items
Invoice-style tables still remap to date, SKU, description, quantity, amount.
Review list
Each run can be saved in this browser on /ai-review. Nothing is stored on our servers.
API
Pro and Team can POST instruction + text to /api/v1/ai and get the same field shape.
Structured data from documents you already upload
Universal extract is a schema on top of the named-field job. It does not certify invoices or read handwriting perfectly.
Confidence is a local score (amount/date/id shape), not a model probability.
For photos of receipts, use Receipt to Excel. For bank PDFs, use Bank statement to Excel.