Getting Started

Requirements

• Node.js ≥ 22.13 (a pdfjs 6 requirement).
• No system dependencies: the canvas implementation (@napi-rs/canvas) and all font / cmap / wasm assets ship as prebuilt artifacts inside the package.

Install / run

Run it directly with npx (no pre-install needed):

npx -y @ai-zen/pdf-parse-mcp

Or install it as a dependency to use the programmatic API:

npm install @ai-zen/pdf-parse-mcp

Wiring an MCP client

Add this snippet to any MCP client (Claude Desktop, Cline, Cursor, your own agent…):

{
  "mcpServers": {
    "pdf-parse": {
      "command": "npx",
      "args": ["-y", "@ai-zen/pdf-parse-mcp"]
    }
  }
}

Once wired, the client can call the three tools pdf_info / pdf_extract / pdf_render.

Your first calls

Given a design.pdf, a typical three-step flow:

1. pdf_info      { "path": "design.pdf" }
                 → how many pages, how big each page is

2. pdf_extract   { "path": "design.pdf", "page": 1 }
                 → all text / shapes / bitmaps on page 1 (detail=box by default, tiny output)

3. pdf_render    { "path": "design.pdf", "page": 1, "scale": 1 }
                 → a full-page thumbnail to eyeball the layout

Coordinate convention

All coordinates use a top-left origin, y down, in points (1pt = 1/72 inch). See Core Concepts.

Next steps

• Full parameters for each tool: MCP Tools
• The complete reconstruction flow: Reconstructing a Design from a PDF