Extract Data from PDFs

You have 50 PDF invoices. You need the amounts, dates, and vendor names in a spreadsheet. Doing this manually takes hours. Claw does it in minutes.

---

What You’ll Accomplish

By the end of this guide, you’ll know how to:

• 📄 Extract structured data from PDFs
• 📊 Export to CSV, Excel, or directly to apps like Notion/Airtable
• 🔍 Parse contracts for key terms and deliverables
• ✅ Verify extracted data before committing

---

Step 1: Point Claw to Your PDFs

Open Claw and describe what you need:

I have 50 invoices in my Documents/Invoices folder.
Extract the vendor name, invoice number, date, and total amount from each.
Export to a CSV file.

Ishi will read your request and ask any clarifying questions.

---

Step 2: Preview the Extraction

Before processing all files, Claw shows a sample extraction:

📄 Sample: Acme-Corp-Invoice-001.pdf
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Extracted Fields:
┌─────────────┬─────────────────────────┐
│ Vendor      │ Acme Corporation        │
│ Invoice #   │ INV-2025-0042           │
│ Date        │ 2025-01-02              │
│ Amount      │ $1,250.00               │
│ Due Date    │ 2025-02-01              │
└─────────────┴─────────────────────────┘

Does this look correct? [Yes, proceed with all] [Adjust fields] [Cancel]

This preview ensures Claw understood your document structure correctly.

---

Step 3: Process All Files

Once you approve the sample, Claw processes all PDFs:

Processing invoices...
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

✅ 48/50 extracted successfully
⚠️ 2 need review (unclear vendor name)

[Review flagged items] [Download CSV]

Files that couldn’t be parsed confidently are flagged for manual review—Claw never guesses.

---

Step 4: Review & Export

Claw shows you the complete dataset before creating any files:

Vendor	Invoice #	Date	Amount
Acme Corp	INV-2025-0042	2025-01-02	$1,250.00
Beta LLC	INV-8821	2025-01-03	$840.00
…	…	…	…

Then choose your export format:

• CSV — Universal spreadsheet format
• Excel (.xlsx) — With formatting
• Direct to Notion — Via MCP integration
• Direct to Airtable — Via MCP integration

---

Example Use Cases

Invoices

Extract vendor, invoice number, date, amount, and due date
from all PDFs in Downloads/Invoices

Contracts

Read this contract and create a list of all deliverables,
deadlines, and payment milestones

Receipts

Parse these expense receipts and create an expense report
with categories, dates, and amounts

Statements

Extract all transactions from this bank statement PDF
and export to Excel

---

Working with Scanned PDFs

For scanned documents (images, not text), Ishi uses OCR:

These are scanned invoices. Use OCR to extract the text first,
then parse the vendor and amounts.

Scanned documents take longer but work with the same workflow.

---

Connecting to Cloud Apps

Want data to go directly to Notion, Airtable, or Google Sheets?

1. Set up the MCP integration for your app
2. Ask Claw to export directly:

Extract invoice data and add each row to my Airtable "Invoices" table

---

Safety Features

Preview Before Commit

Every extraction is previewed before any files are created.

No Guessing

If Ishi can’t confidently parse a field, it flags it for review.

Local Processing

Your PDFs never leave your computer. Extraction happens locally.

---

Tips for Best Results

1. Be specific about field names — “Extract ‘Total Due’ not ‘Subtotal’”
2. Provide a sample — “The vendor name is usually in the top-left header”
3. Start with one file — Test the extraction before running on hundreds

---

Next Steps

• Organize Downloads — Auto-sort files as they arrive
• Cloud Integrations — Connect to Notion, Airtable, etc.
• Batch Processing — Work with hundreds of files