Text Extractor

Lightweight MCP server + CLI for extracting text from PDFs and images, designed for agent workflows.

It is fast on normal digital PDFs, but can still recover text from difficult documents with quality-aware fallback and optional OCR.

Features

• Zero-config extraction for standard PDFs
• Smart fallback chain for low-quality or scanned pages
• Optional OCR stack for image-heavy PDFs and image files
• MCP tools ready for VS Code, Claude, and agent runtimes
• CLI and Python API in one package

Install

Install the package:

pip install text-extractor-lightweight

Optional extras:

# Fast extraction via PyMuPDF (recommended — handles complex fonts, 10-20x faster)
pip install "text-extractor-lightweight[fast]"

# OCR support (scanned PDFs, images)
pip install "text-extractor-lightweight[ocr]"

# Better handling of complex layouts
pip install "text-extractor-lightweight[docling]"

# Everything
pip install "text-extractor-lightweight[all]"

System dependencies for OCR:

• Tesseract OCR: https://github.com/tesseract-ocr/tesseract
• Poppler (pdftoppm) for PDF to image conversion

Windows (winget):

winget install --id tesseract-ocr.tesseract -e
winget install --id oschwartz10612.Poppler -e

CLI

The package name is text-extractor-lightweight, and it installs CLI commands text-extractor and text-extractor-mcp.

# Extract full text from a PDF
text-extractor report.pdf

# Extract from an image
text-extractor screenshot.png

# Extract only specific pages
text-extractor report.pdf --pages 1-5

# Show document metadata/summary only
text-extractor report.pdf --info

# Show which strategy would be used
text-extractor report.pdf --strategy

# Chunk large output by token budget
text-extractor large.pdf --chunk-tokens 50000

MCP Server

Run directly with uvx:

uvx --from text-extractor-lightweight text-extractor-mcp

Claude Code

claude mcp add text-extractor -- uvx --from text-extractor-lightweight text-extractor-mcp

VS Code / GitHub Copilot

Add to .vscode/mcp.json:

{
  "servers": {
    "text-extractor": {
      "command": "uvx",
      "args": ["--from", "text-extractor-lightweight", "text-extractor-mcp"]
    }
  }
}

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "text-extractor": {
      "command": "uvx",
      "args": ["--from", "text-extractor-lightweight", "text-extractor-mcp"]
    }
  }
}

Exposed MCP Tools

Tool	Description
extract_text_from_file	Extract full text from a PDF or image as markdown
extract_text_pages	Extract text from a specific page range
get_document_info	Get page count, type, metadata, and token estimate

Extraction Strategy

Automatic routing selects the best backend by file type and quality:

Image file        -> Tesseract OCR
Digital PDF       -> pymupdf (fast path, recommended)
                     pypdf (fallback if pymupdf not installed)
Garbled PDF text  -> pdfplumber -> docling (optional)
Scanned/image PDF -> pdf2image + Tesseract OCR

PDF fallback chain:

pymupdf -> pdfplumber -> docling -> pdf2image+OCR

PyMuPDF (pip install "text-extractor-lightweight[fast]") is the preferred PDF backend. It correctly decodes custom embedded fonts (where pypdf/pdfplumber emit garbled (cid:N) output) and is 10–20x faster than pypdf.

Python API

from text_extractor.extract import extract_text, extract_raw

# Markdown output
markdown = extract_text("report.pdf")

# Structured output
result = extract_raw("report.pdf")
print(result.total_pages, result.estimated_tokens)
for page in result.pages:
    print(f"Page {page.page_number}: {page.char_count} chars")

Troubleshooting

• If text-extractor is not found on Windows, ensure the Python Scripts directory is in PATH.
• For MCP stdio mode, do not send random JSON to stdin; only an MCP client should talk to the server.
• If OCR is not triggered on scanned docs, confirm Tesseract and Poppler are installed and visible to the process.

Release

Quick publish flow:

python -m build
python -m twine check dist/*
python -m twine upload --repository testpypi dist/*
python -m twine upload dist/*

For the full step-by-step process, see RELEASE_CHECKLIST.md.

License

MIT