pdf-reader

name	pdf-reader
description	Extract text from any PDF — native text via poppler or scanned/image-based via tesseract OCR fallback. Use when the user wants to read, search, summarise, or extract content from a PDF file. This plugin is read-only. For PDF creation, merging, splitting, watermarking, or form filling, use office-pdf instead.

PDF Reader

Extract text from any PDF using pdf-text. Handles both native text PDFs (via pdftotext) and scanned/image-based PDFs (via tesseract OCR fallback). Auto-detects which method to use.

Boundary: Read vs Write

Task	Plugin
Extract text, summarise, search within a PDF	pdf-reader (this plugin)
Create, merge, split, watermark, fill forms	office-pdf

When to Use

• User provides a PDF and wants the text content
• Summarising or searching within a PDF
• Extracting data from scanned documents or invoices
• Checking whether a PDF is text-based or image-based

Usage

# Extract all text (auto-detects text vs image)
pdf-text document.pdf

# Extract specific pages
pdf-text --pages 1-5 document.pdf

# Force OCR even on text-based PDFs
pdf-text --ocr document.pdf

# Show metadata and text/image classification
pdf-text --info document.pdf

Workflow

1. If the user provides a URL, download the PDF first:

   curl -sL -o $CLAUDE_TEMP_DIR/document.pdf "URL"
   

2. Run pdf-text to extract content:

   pdf-text $CLAUDE_TEMP_DIR/document.pdf
   

3. For large PDFs, extract in page ranges to avoid overwhelming context:

   pdf-text --pages 1-10 large-document.pdf
   

4. If you only need metadata or want to check the classification:

   pdf-text --info document.pdf
   

Auto-Detection Logic

The tool samples the first few pages with pdftotext. If fewer than 50 non-whitespace characters are found, it classifies the PDF as image-based and falls back to OCR via tesseract at 300 DPI.

Flags

Flag	Purpose
--pages N-M	Extract pages N through M (inclusive, 1-based)
--pages N	Extract a single page
--ocr	Force OCR regardless of detected text
--info	Show metadata and text/image classification
--version	Show version

Prerequisites

• poppler — provides pdftotext, pdfinfo, pdftoppm (required)
• tesseract — OCR engine (required only for scanned/image-based PDFs)

Install per platform (macOS: brew install poppler tesseract; Debian/Ubuntu: apt install poppler-utils tesseract-ocr; Fedora/RHEL: dnf install poppler-utils tesseract). See plugin README for Arch and Windows.

Common Mistakes

• Using office-pdf to read text from a PDF — use this plugin instead
• Reading the entire 200-page PDF at once — use --pages to extract in manageable chunks
• Assuming OCR failed when output looks rough — scanned PDFs vary in quality; try --info first