XDumps reads the messy PDFs and phone-photo scans nobody wants to retype — invoices, bills of lading, receipts, rate confirmations — and hands back clean, structured data in seconds.
Three steps. Under thirty seconds per document. No template setup, no "training period."
Drag in a PDF or a photo of the document — crooked scans, coffee stains, and fax artifacts included. If a human can read it, XDumps can.
A vision-capable AI model reads the document like a person would and extracts vendor, date, totals, and every line item into a strict, validated schema.
Get clean JSON and a readable table instantly. Paste it into your TMS, spreadsheet, or accounting system — or wire the API straight into your workflow.
If part of someone's job is retyping documents into a system, XDumps pays for itself the first week.
Start with the self-hosted toolkit today, or get on the hosted pilot where we run everything for you.
Tell us what you're drowning in. We'll follow up within one business day — usually with your first documents already extracted.
Straight answers.
No. OCR gives you a blob of text and leaves the hard part — figuring out which text is the vendor, the total, or line 4's unit price — to you. XDumps uses a vision AI model that reads the document the way a person does and returns validated, structured fields. It handles crooked scans, photos, and layouts it has never seen before, with no templates to configure.
On clean, legible documents, field-level accuracy is very high — and every extraction shows you the result next to the source so you can verify in seconds instead of retyping for minutes. For anything the model can't find, you get an explicit null, never a guess dressed up as an answer.
With the self-hosted toolkit, files stay on your machine and are sent only to Anthropic's API for extraction under their commercial data terms (not used for training). With the hosted pilot, documents are processed and retained only as long as you need the results.
A computer with Python (free, and the setup guide walks you through installing it), plus an Anthropic API key. Typical cost per document is one to three cents. If you can install a printer, you can install this — and if you get stuck, email us.
Today you get clean JSON and a copy-paste-ready table; the API endpoint means anything with an import or a Zapier hook can consume it. Direct integrations are prioritized by what pilot customers ask for — tell us what you need in the form above.
Because that's what your document pile is — a dump. We're not going to pretend paperwork is a "digital transformation journey." Dump the doc, keep the data, get back to work.