PDF to Excel: Invoice & Table Extractor

Turn PDF invoices and PDF tables into rows you can open in Excel, Google Sheets or CSV, or send to n8n, Make, Zapier and your database.

• Invoices, bills, purchase orders: supplier, invoice number, invoice date, due date, PO number, customer, VAT/tax ID, currency, subtotal, tax rate, tax amount, total, IBAN, and one row per line item (description, SKU, quantity, unit, unit price, VAT, amount).
• Any other PDF (price lists, statistical reports, schedules, statements): every table becomes rows, with the column names from the table header.
• No AI model and no API key. The PDF's text layer is read with open-source tools (pdfplumber), so results are repeatable and your documents are not sent to any AI service.
• Totals are cross-checked: each line is checked (quantity × unit price = amount) and the lines are checked against the subtotal and total, so you see at once which invoices need a human look.
• $6 per 1,000 PDF pages with data. Pages without any table, failed downloads and scanned pages are free.

What you get

Invoices: one row per line item

Real output for the sample invoice invoice-us-ruled.pdf (a fictional invoice with a bordered item table):

{
  "fileName": "invoice-us-ruled.pdf",
  "documentType": "invoice",
  "rowType": "invoice-line",
  "page": 1,
  "lineIndex": 1,
  "description": "Copy paper A4, 80 g, box of 5 reams",
  "sku": "PAP-A4-80",
  "quantity": 12,
  "unitPrice": 24.9,
  "lineTotal": 298.8,
  "lineCheck": "ok",
  "invoiceNumber": "INV-2026-0142",
  "invoiceDate": "2026-09-15",
  "dueDate": "2026-10-15",
  "poNumber": "PO-7781",
  "supplierName": "Larkspur Sample Supply Co.",
  "supplierTaxId": "00-0000142",
  "customerName": "Bluebird Sample Bakery LLC",
  "currency": "USD",
  "subtotal": 664.17,
  "taxRate": 8,
  "taxAmount": 53.13,
  "total": 717.3,
  "linesSum": 664.17,
  "totalsCheck": "ok",
  "status": "ok"
}

European invoices work too: decimal commas (1.150,00), dates like 30.09.2026, VAT per line, borderless tables and item tables that continue on the next page. Real output for line 4 of the 2-page sample invoice-eu-2-pages.pdf:

{
  "page": 1,
  "lineIndex": 4,
  "description": "Content updates",
  "quantity": 6.5,
  "unit": "h",
  "unitPrice": 42,
  "lineTaxRate": 21,
  "lineTotal": 273,
  "lineCheck": "ok",
  "invoiceNumber": "2026/0387",
  "invoiceDate": "2026-09-30",
  "dueDate": "2026-10-14",
  "supplierName": "Wrenfield Sample Web Studio B.V.",
  "supplierTaxId": "NL000000000B01",
  "customerName": "Harbor Lane Sample Clinic",
  "currency": "EUR",
  "subtotal": 3800.9,
  "taxRate": 21,
  "taxAmount": 798.19,
  "total": 4599.09,
  "totalsCheck": "ok",
  "iban": "NL00EXMP0000000000"
}

All 14 lines of that invoice (9 on page 1, 5 on page 2) come out as 14 rows, and their sum (3,800.90) matches the subtotal.

Any PDF: one row per table row

Real output for the USDA Weekly Combined Regional Shell Egg Report (a public US government price report, 2 pages, 7 borderless tables):

{
  "fileName": "usda-shell-eggs-2026-10-02.pdf",
  "documentType": "tables",
  "rowType": "table-row",
  "page": 1,
  "tableIndex": 1,
  "tableTitle": "National Shell Eggs - Caged · Delivered Warehouse, White, Cents Per Dozen",
  "hasHeader": true,
  "rowIndex": 1,
  "data": {
    "Class": "Extra Large",
    "Price Range": "61.00 - 94.00",
    "Average Price": 81.88,
    "Price Change": 1,
    "Last Reported (9/25/2026)": 80.88
  }
}

Export the dataset to CSV or Excel and every key of data becomes its own column. The 34-page USDA retail specialty crops report gives 1,099 rows from 32 tables in one run.

One Excel file for the whole run

With Save one Excel file on (default), the run also saves OUTPUT.xlsx in its key-value store: a sheet Invoice lines, a sheet Invoices (one row per invoice with the totals and checks) and one sheet per table. The link is in the run's Output tab and in the OUTPUT summary record.

How to use it

1. Paste links to your PDFs into PDF URLs. Google Drive, Dropbox, OneDrive/SharePoint, Box and GitHub share links work (share them as Anyone with the link). You can also use files uploaded to a key-value store, or a dataset of PDF links from a crawler.
2. Keep What to extract = Auto, or choose Invoices or Tables.
3. Click Start. Download the results as Excel, CSV or JSON, open OUTPUT.xlsx, or read them through the API.

Input example

{
  "urls": [
    { "url": "https://raw.githubusercontent.com/Egrarobo/apify-actors/main/pdf-tables-to-rows/samples/invoice-us-ruled.pdf" },
    { "url": "https://www.ams.usda.gov/mnreports/ams_2848.pdf" }
  ],
  "mode": "auto",
  "tableDetection": "auto",
  "parseNumbers": true,
  "dateOrder": "auto",
  "saveExcel": true
}

Field	What it does
urls	PDF links (share links are converted automatically)
kvStoreFileNames, kvStoreId	PDFs uploaded to an Apify key-value store, as "storeName/file.pdf" or "file.pdf" + kvStoreId
datasetId, datasetUrlField	PDF links taken from another Actor's dataset
mode	auto (default), invoice or tables
tableDetection	auto (tables with borders, then borderless), lines or text
parseNumbers	true: 1,234.50, 1.234,50, $12 and (15.00) become numbers. Percentages and codes with leading zeros stay text
dateOrder	auto, DMY or MDY, only for ambiguous dates like 03/04/2026. Invoice dates are returned as YYYY-MM-DD
pdfPassword	Password for protected PDFs
maxPagesPerDocument, maxDocuments, maxFileSizeMb	Limits (0 pages = all)

Output fields

Field	Meaning
status	ok, no-data (no table or invoice found, not charged) or failed (with error, not charged)
rowType	invoice-line, invoice (invoice fields found but no item table) or table-row
lineCheck	ok when quantity × unit price = amount (discounts allowed), mismatch, or computed when the amount was missing
totalsCheck	ok when the lines add up to the subtotal (or total) and subtotal + tax = total; mismatch means: look at this invoice
otherColumns	item-table columns that are not one of the standard fields
tableTitle	the line(s) printed right above a borderless table
tableMethod	lines (table with cell borders) or text (aligned columns)

The OUTPUT record has the run summary: PDFs with data, without data and failed, rows, pages charged, the Excel link and all errors.

Pricing

Pay per page with data.

Event	Price
PDF page with data	$0.006 per page ($6 per 1,000 pages) on the Free plan; $5.50 on Starter, $5 on Scale, $4.50 on Business

A page is charged only if it produced at least one row (an invoice line, an invoice field or a table row). Cover pages, terms and conditions, scanned pages, failed downloads and files that are not PDFs are free. A 1-page invoice costs $0.006; 1,000 two-page invoices cost about $12.

Set a maximum cost per run and the Actor reads only the pages that fit, then stops cleanly. Max pages per PDF caps long files.

Use it in n8n, Make or Zapier

Call the Actor with an HTTP request and get the rows back in the same call:

curl -X POST "https://api.apify.com/v2/acts/egra_van~pdf-invoice-table-extractor/run-sync-get-dataset-items" \
  -H "Authorization: Bearer $APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"urls": [{"url": "https://raw.githubusercontent.com/Egrarobo/apify-actors/main/pdf-tables-to-rows/samples/invoice-eu-2-pages.pdf"}], "mode": "invoice"}'

Typical flow: new invoice PDF in Gmail or Google Drive → this Actor → append the rows to Google Sheets or your accounting tool, and send invoices with totalsCheck = mismatch to a person.

AI agents (Claude, ChatGPT, Cursor, n8n AI Agent) can use it through the Apify MCP server: https://mcp.apify.com?tools=egra_van/pdf-invoice-table-extractor.

Limits, said plainly

• Scanned PDFs (photos or scans without a text layer) are not read. The Actor reports them as no-data and does not charge. If you can, export the PDF from the original program, or OCR it first.
• Invoice fields are found next to their usual labels in English, Romanian, German, French, Dutch, Spanish and Italian (Invoice No, Factura nr, Rechnungsnummer, Bill to, Total due, TVA, MwSt...). An unusual layout may leave some fields empty; the line items and table rows still come out.
• The supplier name is taken from a label (From, Supplier, Furnizor...) or, if there is none, from the largest text at the top of page 1.
• Very complex tables (cells merged over several rows and columns) can come out with col_1, col_2... names instead of the header text. Check hasHeader.

Privacy and personal data

Invoices often contain personal data: names, addresses, tax IDs, bank accounts (IBAN). Here is exactly what happens to them:

• Where the file goes: the PDF is downloaded from your link (or read from your key-value store) into your own Apify run and read there. It is not sent to any AI service or other third party, and the Actor keeps no copy outside your run.
• Where the results stay: the rows, OUTPUT.xlsx and the run log (file names and invoice numbers) are saved in your Apify account, under your account's data retention settings. Delete the run's storage when you no longer need it.
• The Excel link in OUTPUT opens the file without logging in for anyone who has it. Do not post it publicly.
• Share links: a link set to Anyone with the link can be opened by anyone who gets it. For confidential invoices, upload the PDFs to a key-value store and use kvStoreFileNames instead.
• The developer of this Actor does not see your files or results, unless you turn on Share run data with developers in your Apify settings.
• You decide which documents you process; make sure you are allowed to process the personal data in them (for example under the GDPR).

FAQ

Is my data stored or used for AI training? No AI service sees your files, and they are not used for training. They are downloaded into your own Apify run; the results stay in your Apify storage under your retention settings (see Privacy and personal data above).

Can it read Word or Excel files? No, only PDF. For Word, Excel and PowerPoint, use our Document to Markdown Actor.

Why are some numbers text? Percentages (21%), ranges (61.00 - 94.00) and codes with leading zeros (00123) are kept as written. Turn off parseNumbers to keep every cell exactly as printed.

Where do the sample PDFs come from? The two invoices are fictional samples made for this Actor (public domain, CC0). The egg price and petroleum reports are public US government documents (USDA AMS and EIA).