home / mcp / pdf reader mcp server

PDF Reader MCP Server

Provides a production-ready MCP server to extract text, images, and metadata from PDFs via parallel processing across multiple sources.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "sylphlab-pdf-reader-mcp": {
      "command": "npx",
      "args": [
        "@sylphx/pdf-reader-mcp"
      ]
    }
  }
}

You can run a production-ready PDF processing MCP server that lets AI agents extract text, images, and metadata from PDFs with fast, parallel processing and reliable per-page error handling. This server focuses on preserving document layout through Y-coordinate ordering and provides a clean, type-safe TypeScript interface for integration.

How to use

Start the PDF Reader MCP server locally using your MCP client. You will run it as a standard MCP endpoint that your client can connect to for batch and page-level PDF extraction. Use the server to submit sources (local files or HTTP URLs), request full text, metadata, page counts, and embedded images. The tool handles multiple PDFs in parallel and returns results in a natural reading order to support AI agents and downstream processing.

Typical usage patterns include processing a single PDF or batching several PDFs at once. You can request specific pages or ranges, enable image extraction with complete metadata, and obtain per-page content in Y-coordinate order. The server is designed to be easy to integrate with your existing MCP client, with a single tool that handles all operations and strict TypeScript support for reliable development.

If you are building an integration, consider configuring your MCP client to connect to the local stdio endpoint provided by your runtime, using the command and arguments shown in the configuration example. You can also adjust whether you want full text, metadata, page counts, and images according to your use case. The result set includes textual content, images encoded in base64 with width/height/format, and PDF metadata when requested.

How to install

Prerequisites you need before installing: Node.js version 22 or newer. A package manager such as pnpm is recommended, but npm works as well.

Step 1: Install the MCP server locally using the runtime command shown below. You can run without full installation using npx, or install globally if you prefer.

# Quick start - run directly without prior installation
npx @sylphx/pdf-reader-mcp
```

```json
{
  "mcpServers": {
    "pdf_reader": {
      "command": "npx",
      "args": ["@sylphx/pdf-reader-mcp"]
    }
  }
}

Step 2: If you want a persistent setup, install the package globally and configure your MCP client to start the server from a stable command.

npm install -g @sylphx/pdf-reader-mcp

Additional content and notes

Configuration and usage details include various ways to run the MCP server with different clients (Claude Code, VS Code, Cursor, Windsurf, Cline, Warp, Smithery). The common execution pattern is to invoke the MCP server via a command like npx @sylphx/pdf-reader-mcp from your MCP client, ensuring the client can spawn and manage the process.

Recovery and troubleshooting tips cover common issues such as absolute path handling, file-not-found errors, and ensuring the correct working directory. If you encounter a problem with tools not appearing, refreshing the MCP installation and reloading the client typically resolves the issue.

For best results, use absolute paths when referencing local PDFs, enable per-page error isolation, and leverage the Y-coordinate ordering to preserve natural reading flow in AI-driven processing.

Available tools

read_pdf

The single tool that handles all PDF operations.