Documents in, structured JSON out. You name the fields, one API call returns them typed.
Fieldcast is an LLM-native document extraction API. Instead of computer-vision templates or a fixed invoice schema, it reads the document and returns exactly the fields you asked for. You pass a list of field names with the request, up to 40 per call, named however your own database already names them, so a layout nobody has seen before needs no new configuration. The constraint that matters most: it never invents a value. A field absent from the source comes back null, which is what makes it safe for accounting work. An x402 endpoint is live on Base Sepolia for agents that pay per call. Free tier: 50 extractions a month, no card.
-
No templates to configure - field names travel with each request
-
Returns null for missing fields instead of fabricating a value
-
Typed output: numbers as numbers, dates normalised to ISO-8601
-
Up to 40 fields per call, named to match your own schema
-
x402 endpoint so agents can pay per call without an account
-
Extracting supplier, totals, VAT and dates from invoices into accounting systems
-
Pulling merchant, amount and date from expense receipts
-
Reading effective dates, counterparties and notice periods from contracts
-
Giving an AI agent a one-call tool to read a document without loading it into context
-
Agencies onboarding a new client by changing a field list rather than building a parser