✛ Visual Recognition
Document Reader
Send a PNG, JPEG or WebP image, or a PDF with the page you want, by link or as base64. Choose text for everything on the page, table for a table as HTML, or fields with a list of names (invoice number, total, IBAN, anything) to get exactly those back as JSON. One page per call, usually a few seconds. The model runs on our own hardware in the EU: documents are read in memory, never stored and never sent to anyone else. A model can misread; check amounts before you rely on them.
Read one page from an image or PDF. mode=text returns the text, mode=table returns the table as HTML, mode=fields fills in the JSON fields you list.
POST/v1/documents/read
Try it ↓- Price
- $0.0028 per page · REST and MCP
- RapidAPI
- $0.0035 per page · listing at launch
- Returns
- application/json
- MCP tool
- document_reader_read
Request body
| Field | Type | Description |
|---|---|---|
| file_url | string | Direct link to a PNG, JPEG, WebP or PDF file, up to 15 MB. |
| file_base64 | string | The file itself, base64, when it fits in the 2 MB request limit. |
| mode | text | table | fields | — |
| fields | object | For mode=fields: the names to fill in, as an object with empty values, such as
{"invoice_number": "", "total_text": ""}. Up to 40 names; an object may nest, and a
list with one example asks for every such thing on the page. End a name in _text to
get a number exactly as printed: asked for a plain total, the model can read
"15.172,50" as 15.1725. |
| page | integer | Which page of a PDF to read, from 1. The first if left out. |
example
curl https://readymadeapi.com/v1/documents/read \
-H "Authorization: Bearer $READYMADE_KEY" \
-H "Content-Type: application/json" \
-d '{
"fields": {
"currency": "",
"invoice_number": "",
"issue_date": "",
"seller_vat_id": "",
"total_text": ""
},
"file_url": "https://readymadeapi.com/samples/invoice.png",
"mode": "fields"
}'What comes back
{
"mode": "fields",
"fields": {
"currency": "EUR",
"invoice_number": "2026-104",
"issue_date": "2026-11-19",
"seller_vat_id": "RO12345678",
"total_text": "15.172,50"
}
}Use it from an agent
Every endpoint above is also an MCP tool. Add the server to Claude, Cursor or any MCP client and the tools appear with their prices; calls draw from the same credit.
mcp config
{
"mcpServers": {
"readymade": {
"type": "http",
"url": "https://readymadeapi.com/mcp",
"headers": {
"Authorization": "Bearer ${READYMADE_KEY}"
}
}
}
}