Welcome to Ztract — and what we're building

If you’ve ever watched an accounts-payable clerk type vendor names, dates, and line items into a spreadsheet, you already know why we built Ztract. The technology to read those documents has been good enough for years. What was missing was a tool a non-engineer could actually stand up in an afternoon — and a vendor who wouldn’t bury the bill in surprise charges.

We launched the first version of Ztract last week. This is what’s in it, what’s not, and what we’re working on next.

What’s in the box today

You upload a document — PDF, Office file, image, scan, or phone photo. You define the fields you want extracted three ways:

• Pick from one of our ready-made schemas for invoices, receipts, contracts, IDs, resumes, lab reports, customs paperwork, and a handful of others.
• Describe the fields in plain English: “For each invoice, extract the invoice number, vendor, total, and line items with description, quantity, and price.”
• Drop in a sample document and let the engine infer a schema from it — useful when you don’t know what fields exist until you’ve seen one.

You get back structured JSON, CSV, or Excel, with every extracted value anchored to its position on the source document. If anything is off, you fix it in one click. We don’t charge for that — only extraction-time costs pages, not the editing work that comes after.

What’s not in the box yet

We’re dashboard-first. There’s no HTTP API yet, no webhooks, no Zapier connector. Those are on the roadmap and we’ll write about them when they ship. Multi-seat team workspaces are also on the roadmap — each Ztract account today is a single user.

We don’t have a sales team. There’s one inbox — support@ztract.com — and one team behind it. If you want to chat about volume pricing, an enterprise arrangement, or a feature we’re not building, that’s the address.

Pricing, briefly

You buy a one-time page pack. You use the pages over the next 12 months. Nothing recurs and nothing renews. If you don’t use it, you’d get billed nothing. New accounts get 30 free pages to try it out, no credit card.

We’re not running a “free for 14 days then we trap you in a subscription” model. The price is on the pricing page in plain numbers and the smallest pack starts at under $20.

A note on your data

We don’t train models on documents you upload. Not our own engine, not the third-party LLMs we route through. The commercial APIs we use (OpenAI, Anthropic, Google) prohibit training on API-submitted data, and we rely on those commitments. When you delete a document, it’s deleted — immediately from active storage and within 14 days from backups.

If you’d like the formal picture, our Privacy Policy and Data Processing Agreement describe everything in detail.

What we’d love feedback on

Three things, in order of how much it helps us:

1. Documents we get wrong. If a layout we should handle comes back with missing or misplaced values, we’d genuinely love to see it. Email a sample (anonymized if you need) and we’ll dig in.
2. The schema-design flow. We’ve tried hard to make schema creation feel obvious to non-engineers. If something tripped you up, tell us.
3. What’s missing from the product that would make you commit to using it for real volume. We can’t build everything, but we’d rather hear the gap than guess at it.

Either way — thanks for trying it out. There’s a lot more coming.