pdffigures2

pdffigures2

Installation

Install the skill into your agent that supports Agent Skills with skit:

skit install github:vlln/pdffigures-mcp-server/skills/pdffigures2

Or manually copy the skills/pdffigures2 directory into your agent's skills directory, then configure:

cp .env.example .env
# edit PDFFIGURES2_API_URL if the server is remote

When To Use

Use this skill when the user asks to:

• Extract figures, charts, illustrations, or tables from a PDF
• "Get the figures out of this paper"
• "What tables are in this PDF?" or "Extract all charts from this document"
• Understand a PDF's visual content beyond just its text

The skill calls a pdffigures2 extraction server via HTTP. It identifies figure and table regions, extracts captions, runs OCR on table text, and renders each element as a PNG.

Do not use this skill for: extracting only the text layer (use pdf2text), extracting embedded image assets without caption matching (use pymupdf), or processing non-scholarly PDFs like brochures (results may be poor).

Workflow

1. Ensure the server is running

This skill is an API client — it requires a pdffigures2 server.

Point to an existing remote server by setting PDFFIGURES2_API_URL in .env (copy from .env.example).

2. Run extraction

skills/pdffigures2/scripts/pdffigures2 <pdf_path> -o <output_dir>

Options:

• -o, --output-dir DIR — Where to download figures (default: ./pdffigures2_output)
• --quiet — Suppress progress messages

Example:

skills/pdffigures2/scripts/pdffigures2 paper.pdf -o ./extracted/

3. Interpret the output

The script prints a JSON summary to stdout containing:

{
  "total": 5,
  "figures": [
    {
      "figType": "Figure",
      "name": "3",
      "page": 4,
      "caption": "Fig. 3. Overview of the proposed architecture.",
      "captionBoundary": { "x1": 100, "y1": 200, "x2": 400, "y2": 215 },
      "regionBoundary": { "x1": 60, "y1": 80, "x2": 440, "y2": 350 },
      "renderDpi": 300,
      "imageText": [],
      "filePath": "/absolute/path/to/extracted/paper-Figure3-1.png"
    },
    {
      "figType": "Table",
      "name": "I",
      "page": 2,
      "caption": "TABLE I Dataset statistics.",
      "imageText": ["Column A", "Column B", "value 1", "value 2"],
      "filePath": "/absolute/path/to/extracted/paper-TableI-1.png"
    }
  ],
  "figuresDir": "/absolute/path/to/extracted/",
  "elapsedSeconds": 3
}