paper-parse

name	paper-parse
description	Parse academic PDF papers into markdown with figure extraction.
metadata	{"openclaw":{"emoji":"📄","requires":{"bins":["uv"]}}}

Paper Parse

Parse academic PDF papers into structured markdown with figure extraction using PyMuPDF.

Usage

uv run {baseDir}/scripts/parse_paper.py --pdf /path/to/paper.pdf [--output-dir ./output]

Output

The tool generates:

• {paper_name}_content.md - Full paper content in markdown
• {paper_name}_parsed.json - Structured metadata including:
  • Paper title
  • Number of pages
  • Extracted figures with captions and paths
• cover_title_authors.png - First-page snapshot focused on title + authors region
• figures/ - Directory containing high-resolution figure screenshots

Example

uv run scripts/parse_paper.py --pdf ~/papers/my-paper.pdf --output-dir ./parsed

Output structure:

./parsed/
├── my-paper_content.md
├── my-paper_parsed.json
└── figures/
    ├── figure_1.png
    ├── figure_2.png
    └── ...

Dependencies

• PyMuPDF (fitz) - PDF parsing and rendering
• pymupdf4llm - Markdown conversion

These are automatically managed by uv via the inline script metadata.