mcp-server-pdf

A Model Context Protocol (MCP) server that exposes PDF operations — text extraction and metadata retrieval — to MCP-compatible clients such as Claude Desktop.

Features

• Extract text from a PDF (all pages or a specific page)
• Get metadata from a PDF (title, author, page count, creation date, etc.)

Requirements

• Go 1.25.1+
• (Optional) Nix with flakes enabled for a reproducible dev environment

Installation

go install

go install github.com/tkuramot/mcp-server-pdf@latest

Build from source

git clone https://github.com/tkuramot/mcp-server-pdf.git
cd mcp-server-pdf
go build -o mcp-server-pdf .

Using Nix

git clone https://github.com/tkuramot/mcp-server-pdf.git
cd mcp-server-pdf
nix develop
go build -o mcp-server-pdf .

Usage

The server communicates over stdio following the MCP specification. It is not intended to be run directly in a terminal; instead, register it with your MCP client.

Claude Desktop

Add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "pdf": {
      "command": "/path/to/mcp-server-pdf"
    }
  }
}

Replace /path/to/mcp-server-pdf with the absolute path to the binary.

Tools

extract_text

Extracts plain text content from a PDF file.

Example response (all pages):

--- Page 1 ---
Lorem ipsum dolor sit amet...

--- Page 2 ---
Consectetur adipiscing elit...

get_metadata

Returns metadata embedded in a PDF file as JSON.

Example response:

{
  "title": "Annual Report 2024",
  "author": "Jane Doe",
  "subject": "Finance",
  "keywords": ["finance", "annual"],
  "creator": "Microsoft Word",
  "producer": "Adobe PDF Library",
  "creation_date": "2024-01-15T09:00:00Z",
  "modification_date": "2024-03-01T12:00:00Z",
  "page_count": 42
}

Development

# Run tests
go test ./...

# Lint
go vet ./...

With Nix:

nix develop

Dependencies

License

MIT