home / mcp / pdf reader mcp server

PDF Reader MCP Server

Provides production-grade PDF processing with text, image, and metadata extraction from local or URL sources using parallel MCP-based execution.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "sylphxai-pdf-reader-mcp": {
      "url": "https://your-server.com/mcp",
      "headers": {
        "MCP_API_KEY": "YOUR_API_KEY",
        "MCP_HTTP_HOST": "0.0.0.0",
        "MCP_HTTP_PORT": "8080",
        "MCP_TRANSPORT": "http"
      }
    }
  }
}

PDF Reader MCP provides production-grade PDF processing capabilities for AI agents, delivering fast text, image, and metadata extraction with natural content ordering and per-page resilience. It enables scalable, parallel PDF processing for applications that rely on accurate document understanding and rapid ingestion.

How to use

You will run the MCP server locally or remotely and connect your MCP client to send PDF processing requests. Use the single tool to request text, images, metadata, and page counts from one or multiple PDFs. Expect natural document ordering based on Y-coordinates, parallel processing for speed, and per-page error handling that prevents a single failure from stopping the entire batch.

Typical workflows include extracting full text from a document, retrieving per-page metadata, selecting specific pages or ranges, and batching multiple PDFs for concurrent processing. You can also request embedded images with their metadata and receive content in the original reading order for AI models that need spatial context.

How to install

Prerequisites: ensure your environment has Node.js version 22 or newer and a package manager such as pnpm or npm.

Install and run the MCP server locally using the standard CLI approach.

# Quick start: run without installation in your project
npx @sylphx/pdf-reader-mcp

# Or install globally for repeated use
npm install -g @sylphx/pdf-reader-mcp

Additional setup for HTTP transport (remote access)

If you want to expose the MCP server over HTTP so multiple machines can access it, enable the HTTP transport by setting the transport mode and port.

# Run as an HTTP server on port 8080
MCP_TRANSPORT=http npx @sylphx/pdf-reader-mcp

Docker deployment

You can containerize the MCP server and enable HTTP transport within the container. The example shows how to install the package and expose port 8080.

FROM oven/bun:1
WORKDIR /app
RUN bun add @sylphx/pdf-reader-mcp
ENV MCP_TRANSPORT=http
ENV MCP_HTTP_PORT=8080
EXPOSE 8080
CMD ["bun", "node_modules/@sylphx/pdf-reader-mcp/dist/index.js"]

Connecting from an MCP client

When you configure an MCP client, you reference the server as a resource for PDF processing tasks. If you run with HTTP transport, point your client at the server using the provided URL and, if needed, attach authentication via an API key.

Notes on usage patterns

• Absolute and relative paths are supported for local PDFs. Absolute paths are available starting from a recent release, allowing direct file access without path gymnastics. The system can process multiple PDFs in parallel to maximize throughput.

• Contents are ordered by Y-coordinate to preserve reading order, which helps AI understand spatial relationships within documents.

Troubleshooting and tips

If you encounter path-related errors, ensure your working directory is correct or switch to an absolute path. If you need to restart after making changes, fully restart the MCP client.

For remote access, verify that MCP_TRANSPORT is set to http and that the HTTP server port is reachable from your network.

Available tools

read_pdf

The principal tool that handles all PDF operations, accepting a list of sources and flags to extract text, metadata, images, and page counts.