Turn any document into structured JSON in one API call.
You send a PDF or raw text and name the fields you want. Fieldcast returns them as typed JSON — numbers as numbers, dates normalised to ISO. No templates to draw, no training data to label, no model to fine-tune.
Built for invoices, receipts, purchase orders, contracts and statements.
One request, one response
Name the fields. Get them back. That is the entire interface.
curl -X POST https://fieldcast-peach.vercel.app/v1/extract \
-H "X-API-Key: demo_key_public" \
-H "Content-Type: application/json" \
-d '{
"text": "INVOICE INV-2043 ... Date: 14 August 2026 ... Total: 1,240.50 EUR",
"fields": ["invoice_number", "date", "total", "currency"]
}'{
"invoice_number": "INV-2043",
"date": "2026-08-14",
"total": 1240.5,
"currency": "EUR"
}PDFs work the same way — post the file as multipart with fields=a,b,c. Any field the document does not contain comes back null rather than invented.
Why teams pick it
No template drawing
Template-based parsers break the moment a supplier changes their layout. Fieldcast reads the document, so a new layout is not a new project.
Fields you choose
Not a fixed invoice schema. Ask for up to 40 fields, named however your database already names them.
It returns null, not a guess
Missing data comes back as null. A parser that invents a total is worse than one that admits it did not find one.
Typed output
Totals come back as numbers, dates as ISO-8601. No downstream string cleanup.
Pay-per-call for AI agents
Fieldcast also exposes an x402 endpoint: no signup, no API key. An agent calls it, receives HTTP 402 with payment terms, pays in USDC, and gets the extraction back in the same flow. Live on Base Sepolia (a test network) today.
No account needed
An autonomous agent cannot fill in a signup form. It can pay per call.
Settlement in cents
$0.01 per extraction. Running on Base Sepolia, a test network — mainnet is not enabled yet.
Discoverable
Declared to the x402 Bazaar so agents searching for document extraction can find it on their own.
Pricing
| Plan | Extractions | Price |
|---|---|---|
| Free While in beta | 50 / month | $0 |
| Starter | 1,000 / month | $19 /mo |
| Pro | 5,000 / month | $49 /mo |
| Agents (x402) | Unmetered | $0.01 /call · Base Sepolia testnet |
Every plan uses the same endpoint and the same extraction engine. No feature is held back behind a higher tier.
Frequently asked questions
Do I have to define a template for each document layout?
No. You pass a list of field names with the request. Fieldcast reads the document and returns those fields. A supplier changing their invoice layout does not require any change on your side.
What file types are supported?
PDF and plain text today. PDFs are converted to text server-side before extraction. Scanned image-only documents need OCR, which is not yet available.
What happens if a field is missing from the document?
It is returned as null. Fieldcast does not invent values to fill a schema, because a plausible wrong total is more expensive than a missing one.
How accurate is it?
On the reference test — a ten-field invoice — all ten fields were extracted correctly, with dates normalised to ISO-8601 and monetary values returned as numbers. Accuracy on your own documents should be measured on your own documents, which is what the free tier is for.
Can an AI agent use Fieldcast without an account?
Yes. The x402 endpoint requires no API key. The agent receives payment terms as an HTTP 402 response, settles in USDC, and receives the extraction in the same request flow. It runs on Base Sepolia today, a test network, so the payment is not real money yet.
Is my document stored?
Document contents are not persisted. Only a usage counter — API key, month, call count, character count — is recorded for billing.