In validation — public pricing is live

Put useful technology within reach of more people.

Typed, auditable document extraction at the edge.

EdgeExtract is a raw, developer-first API from RedSugarTree that turns invoices, receipts, tickets and emails into typed fields with confidence scores and a missing-probability (p_null) — not fragile LLM JSON. Powered by Cloudflare Clairic, running on the edge. $0.10 per million input tokens, no output fee. Pay per call, not per page.

No credit card to try the Free tier. Real prices, no launch discounts.
Why this exists

Why not just prompt an LLM?

You can ask ChatGPT to "extract the total." But you can't route on its certainty, and you can't prove what it read. EdgeExtract returns structure you can build automation around.

∅

Typed + p_null

Every field returns a valid typed value or null with a probability it is missing. Route the sure ones to automation, the rest to a human.

✓

Audit trail

Confidence scores + character spans mean you can prove what was extracted and why — something free-form prompts can't give you.

$

Cheap & edge-native

$0.10/M input tokens, no output token charge. Runs on Cloudflare Workers — no GPU to host, no bill to babysit.

⤓

Embed once, keep it

Drop in an API call. Switching later means changing your integration code — that's the lock-in that makes it worth paying for.

⟲

Honest about limits

Clairic is a 2026 preview. One- and two-step math is reliable; long chains are not. We tell you where, and p_null shows you the gaps.

⌘

Built by one person

A RedSugarTree tool. Small enough that the developer who built it also answers your email — no ticket queue.

How it works

Send a schema, get typed fields back.

Three steps, then a request you could paste into your terminal today.

1

Send input + schema

POST the document text (or up to 4 images) with a schema of typed fields you want.

2

Get typed values

Clairic returns one typed value per field, each with a confidence score and p_null (missing probability).

3

Route on confidence

Automate the high-confidence fields; send the rest to a human review queue. No guessing.

// POST /api/extract
const res = await fetch("/api/extract", {
  method: "POST",
  body: JSON.stringify({
    state: invoiceText,
    fields: {
      invoice_number: { type: "text" },
      total: { type: "number", decimals: 2, unit: "USD" }
    }
  })
});
// => {
//      invoice_number: "INV-2291",
//      total: 701.10,
//      _meta: { confidence: 0.98, p_null: 0.01 }
//    }
Pricing

Pay per call, not per page.

Real prices, no discounts. A 12-page invoice costs the same per field as a 1-page one — multi-page docs don't punish you.

Free

For testing the API and small projects.

$0/mo
  • 1,000 extractions / month
  • Typed + confidence + p_null
  • Community support

Pro

For teams with higher volume and custom needs.

$29/mo
  • 100,000 extractions / month
  • Custom schemas
  • Priority support

Overage: $0.50 per 10,000 extractions. Hard cap available on request.

Questions

Frequently asked

Is Clairic accurate enough for invoices?

Clairic is a 2026-10 preview from Cloudflare. One- and two-step math (totals, dates) is reliable; long chains are not — Cloudflare recommends extracting inputs with Clairic and doing math in your own code. Use the confidence score to route review, and p_null to catch fields that simply aren't in the document.

How is this different from AWS Textract?

Textract is per-page and cloud-locked. EdgeExtract is input-token priced with no output fee, edge-native, and returns p_null + confidence for review routing out of the box.

How is this different from DigiParser and other AppSumo extraction tools?

Those are UI-heavy workflow builders sold as lifetime deals. Developers who just want a parsing engine find them overkill; they burn page-credits on multi-page PDFs; and per-field confidence + REST API are locked behind their top tiers. EdgeExtract is the opposite: a bare extraction API, confidence and p_null open at every tier, and you pay per call, not per page — so a 12-page invoice costs the same per field as a 1-page one.

Can I self-host?

The extraction model is hosted by Cloudflare. EdgeExtract wraps it into a typed, auditable API so you don't maintain the pipeline, the rate limits, or the 24/7 reliability yourself.

What happens to the documents I send?

Documents are sent to Cloudflare Workers AI (Clairic) for extraction and are not used to train models. Full details are in our Privacy Policy.

Start extracting. Pay only for what you call.

Start free

A RedSugarTree tool — small software, one job at a time.