Upload a PDF, scan, or image. Get JSON, CSV, or Excel in seconds — no model training, no monthly subscription, no setup.
From invoices and contracts to passports and lab reports — across formats, layouts, and languages.
Our document understanding engine reads tables that span pages, stamps over text, and faded thermal receipts — the parts that break traditional OCR.
Volume packages from $19.90. No subscription, no per-seat fees. Pages stay valid for 12 months, so you only pay for the work you actually need.
Describe your fields in plain English, pick a ready-made template, or drop in a sample document — Ztract figures out the schema in a few seconds.
How it works
You don't write code. You don't train a model. You define what you want, then upload.
Group related documents — invoices, contracts, IDs — into one workspace. Each project has its own schema, its own settings, and its own export targets.
Tell the engine which fields to extract from each document type — invoice number, line items, dates, parties, whatever your project needs. Three ways to define it: a ready-made template, a plain-English description, or a sample document.
Drag in PDFs, Office documents, images, or scans — one file or many in the same session. Documents are parsed as they arrive — no queue surprises.
Open the side-by-side viewer to verify every field against the source. Fix anything you don't like in one click. Export to JSON, CSV, or Excel.
An HTTP API, webhooks, and platform integrations are on the roadmap.
Schema design
You don't need to be a developer to create an extraction schema. Pick whichever feels natural — or mix all three.
Pick a ready-made schema for invoices, receipts, contracts, IDs, and more — built from real-world layouts. Rename fields to your team's terminology, save it once, and reuse it across thousands of documents.
Type what you need: "For each invoice, extract the invoice number, total, vendor, and line items with description, quantity, and price." Ztract drafts a complete schema from your description — review and refine before you upload.
Drop in one example document. The engine reads it and proposes a schema with the right field names, types, and nesting. Perfect when you don't know exactly which fields exist until you've seen one — common for medical reports, custom contracts, and one-off vendor forms.
Live demo
Click a field on the right and we'll highlight where it came from on the left. Click a region on the left to jump to its data.
| Date | Debit | Credit | Balance | Description |
|---|---|---|---|---|
| Quantity | Subtotal | Tax rate | Unit price | Description |
|---|---|---|---|---|
| Code | Price | Amount | Discount | Quantity | Description |
|---|---|---|---|---|---|
| Major | Degree | School | End date | Start date |
|---|---|---|---|---|
| Title | Company | End date | Start date | Description |
|---|---|---|---|---|
| Origin | Hs code | Currency | Quantity | Incoterms | Net weight | Unit price | Total weight | Product description |
|---|---|---|---|---|---|---|---|---|
Sample data. Real engine output.
Verify & correct
OCR services hand you a pile of text. Ztract hands you structured data with a built-in viewer that tells you exactly where each value came from — and lets you correct anything that's off, without a re-run.
Click any extracted field and the matching region lights up on the document. Click a region on the document and the field scrolls into view. No more "where did this number come from?" guesswork during audits or internal reviews.
When the engine isn't 100% sure, it tells you. Low-confidence fields are flagged in the viewer so reviewers know exactly which values to spot-check — instead of re-reading 100 fields hoping to catch the one that might be wrong.
Spot a wrong value? Click it, type the right one, save. The correction is instant and flows through to your next export. You don't pay for a re-extraction, and the original bounding box stays attached so the audit trail survives.
Engineered for accuracy
Most parsers either match characters from pixels (and break on anything unusual) or pipe a screenshot to a language model (and hallucinate). We built something else — purpose-built for documents your team actually receives.
Reads what's on the page, not just what's printed in a straight line.
You describe the shape you want — we figure out where to find it.
Built for teams that have to defend the numbers they extract.
The hard problems aren't on the test set. They're in the email attachments your finance team forwards you on a Tuesday.
A new vendor invoice works the first time, without you adding a template or a rule. The engine learns layout from the document itself.
OCR returns characters. We return meaning. Dates, currencies, tax IDs come back normalized — and fields that are implied get extracted, not just the ones that are labeled.
One pipeline for PDFs, Office documents, scans, phone photos, and text-based formats — one schema across all of them.
Use cases
We ship starter schemas and sample documents for the workflows teams ask about most.
Line items, totals, tax IDs, currencies — across thousands of vendor layouts.
Learn more →Merchant, items, tax, tip — even when the paper is wrinkled or faded.
Learn more →Opening and closing balances, every transaction, with running totals.
Learn more →Parties, effective dates, renewal terms, governing law, signatures.
Learn more →Names, numbers, dates, MRZ — with locale-aware field validation.
Learn more →Education, experience, skills — normalized into a hiring-ready shape.
Learn more →Patient info, test panels, reference ranges, abnormal flags.
Learn more →HS codes, weights, ports, parties — for trade compliance teams.
Learn more →Why we built it
“Every team I know wastes a weekend a quarter on data entry that should take ten minutes. We built Ztract so that weekend goes back to you.”
Pricing
Pay as you go. Pick a pack that fits your volume — every pack stays valid for a full year.
100 pages · $0.199 €0.179 £0.149 $0.279 $0.279 ¥0.999 ₩300 ¥32 / page
Valid for 12 months
Buy now500 pages · $0.080 €0.070 £0.060 $0.110 $0.110 ¥0.400 ₩120 ¥14 / page
Valid for 12 months
Buy now1,000 pages · $0.065 €0.060 £0.050 $0.090 $0.100 ¥0.270 ₩100 ¥11 / page
Valid for 12 months
Buy now2,000 pages · $0.055 €0.050 £0.045 $0.080 $0.080 ¥0.350 ₩85 ¥9 / page
Valid for 12 months
Buy now5,000 pages · $0.040 €0.036 £0.030 $0.058 $0.058 ¥0.200 ₩62 ¥7 / page
Valid for 12 months
Buy now10,000 pages · $0.037 €0.033 £0.028 $0.052 $0.052 ¥0.200 ₩56 ¥6 / page
Valid for 12 months
Buy now20,000 pages · $0.035 €0.031 £0.026 $0.050 $0.050 ¥0.200 ₩53 ¥6 / page
Valid for 12 months
Buy now50,000 pages · $0.030 €0.026 £0.022 $0.042 $0.042 ¥0.160 ₩45 ¥5 / page
Valid for 12 months
Buy nowAll packages: 12-month validity · JSON / CSV / Excel export · Side-by-side review
FAQ
Start with 30 free pages. No credit card. No subscription.
