LLM PDF Parser

PDF Parser MCP Server

A Model Context Protocol (MCP) server that extracts text from PDF files, with optional OCR support for scanned documents.

Features

• Text extraction from PDF files using pypdf
• Word Document support - read .docx files with formatting (headers, tables)
• OCR support via Tesseract for scanned/image-based PDFs and DOCX images
• URL support - read documents directly from HTTP/HTTPS URLs
• Page selection - extract specific pages or ranges (PDF only)
• Password protection - decrypt protected PDFs

Prerequisites

For OCR support (highly recommended), install Tesseract OCR:

• Windows: Download from UB-Mannheim or run choco install tesseract
• Default path: C:\Program Files\Tesseract-OCR\tesseract.exe

Installation & Setup

Option 1: Automatic (Windows)

No manual setup required. The included start_server.bat script will automatically create a virtual environment and install all dependencies the first time you run it.

Option 2: Manual (Mac/Linux/Custom)

python -m venv .venv
source .venv/bin/activate  # or .venv\Scripts\activate on Windows
pip install -r requirements.txt

Configuration

1. Gemini CLI

Method A: Using Batch Script (Recommended for Windows)

gemini mcp add pdf-parser "C:\Path\To\LLM-PDF-Parser\start_server.bat"

Method B: Using Python Directly

gemini mcp add pdf-parser "python" --args "C:\Path\To\LLM-PDF-Parser\server.py"

2. Claude Desktop & Other MCP Clients

Add the following to your MCP configuration file (e.g., claude_desktop_config.json or VS Code settings).

Method A: Using Batch Script (Recommended for Windows) Ensures dependencies are managed automatically.

{
  "mcpServers": {
    "pdf-parser": {
      "command": "C:\\Path\\To\\LLM-PDF-Parser\\start_server.bat",
      "args": [],
      "env": {
        "PYTHONIOENCODING": "utf-8"
      }
    }
  }
}

Method B: Using Python Directly Requires you to handle the virtual environment and dependencies manually.

{
  "mcpServers": {
    "pdf-parser": {
      "command": "python",
      "args": ["C:\\Path\\To\\LLM-PDF-Parser\\server.py"],
      "env": {
        "PYTHONIOENCODING": "utf-8"
      }
    }
  }
}

Tool: read_pdf

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | file_path | string | Yes | Absolute path to PDF or URL (http/https) | | pages | string | No | Page selection: "0,1,5" or "0-5" (0-indexed) | | password | string | No | Password for encrypted PDFs | | ocr | boolean | No | Enable Tesseract OCR for images (default: true) |

Tool: read_docx

| Parameter | Type | Required | Description | |-----------|------|----------|-------------| | file_path | string | Yes | Absolute path to DOCX or URL (http/https) | | ocr | boolean | No | Enable Tesseract OCR for embedded images (default: true) |

Examples

# Read entire PDF
read_pdf("C:/docs/report.pdf")

# Read specific pages from PDF
read_pdf("C:/docs/report.pdf", pages="0,2-5")

# Read DOCX file (preserves headers and tables)
read_docx("C:/docs/notes.docx")

# Read from URL
read_pdf("https://example.com/document.pdf")
read_docx("https://example.com/meeting.docx")

# Disable OCR (enabled by default)
read_pdf("C:/docs/scanned.pdf", ocr=False)

# Decrypt protected PDF
read_pdf("C:/docs/protected.pdf", password="secret")

Requirements

• Python 3.10+
• pypdf - PDF text extraction
• python-docx - Word document extraction
• pytesseract - OCR wrapper (optional)
• Pillow - Image processing (optional)
• requests - URL downloads
• mcp - MCP server framework

Notes

• Uses pure Python pypdf instead of pymupdf for ARM64 compatibility
• OCR requires Tesseract to be installed on the system
• Page numbers are 0-indexed in the pages parameter

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Beledarian
• Source: Beledarian/LLM-PDF-Parser
• License: MIT

Install and usage instructions live in the source repository linked above.