# Invoice OCR & Parser

Invoice OCR to JSON: seller, buyer, tax IDs, IBAN, line items and totals, with amounts parsed by fixed rules, not guessed.

Send an invoice as a PNG, JPEG, WebP or PDF, by link or as base64. You get one JSON object with the invoice number, issue and due dates, seller and buyer with their tax IDs, every line item, subtotal, tax, total, currency and IBAN. Amounts are read exactly as printed and then turned into numbers by fixed rules, so 15.172,50 and 15,172.50 both come out as 15172.5; the printed text comes back beside each number so you can check it. Up to three pages per call, a few seconds each. Runs on our own hardware in the EU; invoices are read in memory and never stored. A model can misread: check totals before you pay on them.

- Amounts are read as printed and turned into numbers by fixed rules, so 15.172,50 and 15,172.50 both give 15172.5. The printed text comes back beside each number.
- Invoice data extraction into one JSON object: number, dates, seller and buyer with tax IDs, every line item, subtotal, tax, total, currency and IBAN.
- Invoices are read in memory on our own hardware in the EU and never stored.

## Calls

### `POST https://readymadeapi.com/v1/invoices/parse`

Send an invoice and get its number, dates, seller, buyer, line items, subtotal, tax, total, currency and IBAN as JSON. Each page read is one unit.

- Price: $0.01 per page, up to 3 per call
- MCP tool: `invoice_parser_parse`
- Answers in the same call.

Request:

```json
{
  "file_url": "https://readymadeapi.com/samples/invoice.png"
}
```

Answer:

```json
{
  "fields": {
    "invoice_number": "2026-104",
    "issue_date": "2026-11-19",
    "due_date": "2026-12-19",
    "purchase_order": "4500012345",
    "currency": "EUR",
    "seller": {
      "name": "Northwind Traders SRL",
      "address": "Str. Exemplu 1, 400000 Cluj-Napoca, Romania",
      "tax_id": "RO12345678"
    },
    "buyer": {
      "name": "Client GmbH",
      "address": "Beispielweg 7, 10115 Berlin, Germany",
      "tax_id": "DE811111111"
    },
    "items": [
      {
        "description": "Oak desk 160 cm",
        "quantity": 40.0,
        "unit_price": 180.0,
        "amount": 7200.0
      },
      {
        "description": "Oak shelf, wall mounted",
        "quantity": 65.0,
        "unit_price": 80.0,
        "amount": 5200.0
      },
      {
        "description": "Delivery",
        "quantity": 1.0,
        "unit_price": 350.0,
        "amount": 350.0
      }
    ],
    "subtotal": 12750.0,
    "tax": 2422.5,
    "total": 15172.5,
    "iban": "RO49AAAA1B31007593840000"
  },
  "as_printed": {
    "items.0.quantity": "40",
    "items.0.unit_price": "180,00",
    "items.0.amount": "7.200,00",
    "items.1.quantity": "65",
    "items.1.unit_price": "80,00",
    "items.1.amount": "5.200,00",
    "items.2.quantity": "1",
    "items.2.unit_price": "350,00",
    "items.2.amount": "350,00",
    "subtotal": "12.750,00",
    "tax": "2.422,50",
    "total": "15.172,50"
  },
  "pages_read": 1,
  "pages": 1
}
```

## Limits

- Up to 3 pages per call: PNG, JPEG, WebP or PDF.
- By link, up to 15 MB.

## Questions

**Which files and how many pages?**

PNG, JPEG, WebP or PDF, by link or as base64. Up to three pages per call, a few seconds each; each page read is one unit.

**How accurate is it?**

It is a model and it can misread. That is why each amount comes with the text as printed: check totals before you pay on them.

**Does it read e-invoices such as XRechnung or ZUGFeRD XML?**

No. It reads the page as a person sees it.

## Using it

- Send `Authorization: Bearer <key>` and a JSON body. Keys and credit: https://readymadeapi.com/account
- Every answer carries `X-Request-Id`, `X-Units` and `X-Charged-Usd`.
- An error is a JSON problem with a `title` to act on and a `detail` in plain words; nothing is charged for it.
- Full schema: https://readymadeapi.com/openapi.json
