How to Parse Receipts with an API (Python + Node.js)

Every expense management app has the same problem: users upload photos of receipts, and someone has to manually read the merchant name, total, tax,

tip, and line items. It's tedious, error-prone, and doesn't scale.

I built aPapyr to solve this. It's an API that reads receipts (and invoices, tax forms, bank statements) and returns structured JSON — including per-field confidence scores so you know what to trust.

Here's how to use it in Python and Node.js.

## Python

bash
  pip install apapyr

  from apapyr import aPapyr

  client = aPapyr("sk_live_your_key")
  result = client.extract("receipt.jpg", document_type="receipt")

  print(result.get_field("merchant_name"))  # "Starbucks Coffee"
  print(result.get_field("total"))           # 15.57
  print(result.get_field("tax"))             # 1.12
  print(result.get_field("tip"))             # 2.00
  print(result.get_field("payment_method"))  # "Visa ending 4242"

  Node.js

  npm install apapyr

  const { aPapyr } = require("apapyr");

  const client = new aPapyr("sk_live_your_key");
  const result = await client.extract("receipt.jpg", {
    documentType: "receipt"
  });

  console.log(result.getField("merchant_name")); // "Starbucks Coffee"
  console.log(result.getField("total"));          // 15.57
  console.log(result.getField("tip"));            // 2.00

  What You Get Back

  The API doesn't just OCR the text — it understands the receipt. It knows which number is the total, which is the tax, and which is a line item price.   Every field includes a confidence score:

  print(result.get_field_confidence("total"))          # 0.98 — very confident
  print(result.get_field_confidence("merchant_name"))  # 0.99
  print(result.get_field_confidence("tip"))             # 0.91 — handwritten, slightly less sure

  This lets you build smart automation: auto-process anything above 0.95 confidence, flag the rest for human review.

  Line Items

  It pulls individual items too:

  for item in result.line_items:
      name = item.get("description", {}).get("value")
      price = item.get("amount", {}).get("value")
      print(f"  {name}: ${price}")

  # Grande Caramel Macchiato: $5.95
  # Blueberry Muffin: $3.50
  # Bottled Water: $3.00

  Handles Messy Photos

  Crumpled receipt from your pocket? Blurry phone photo? Faded thermal paper? It uses AI vision models (not old-school OCR), so it understands context.   If the "5" in "$15.57" is smudged, it still knows the total because the line items add up to it.

  Auto-Detect Document Type

  Don't know if the user uploaded a receipt or an invoice? Let the API figure it out:

  result = client.extract("mystery_document.pdf")  # document_type defaults to "auto"
  print(result.document_type)  # "receipt"

  Works With AI Agents

  If you use Claude Code or Cursor, you can skip the SDK entirely:

  claude mcp add apapyr -- npx apapyr-mcp-server

  Then just say: "Parse this receipt and tell me the total." The AI agent calls aPapyr automatically.

  Try It

  - https://apapyr.com/demo.html — see real extraction results with sample documents
  - https://apapyr.com/free-tool.html — upload your own receipt, no signup needed
  - https://apapyr.com/dashboard.html — 50 pages/month free
  - https://apapyr.com/docs.html — full reference

  The free tier is enough to build and test your integration. Paid plans start at $49/month for 1,000 pages.

  ---
  aPapyr is on https://github.com/AkilaJ?tab=repositories&q=apapyr, https://pypi.org/project/apapyr/, and https://www.npmjs.com/package/apapyr.