io.github.agenson-horrowitz/document-parser

Multi-Format Document Parser MCP Server

A professional-grade MCP server that provides AI agents with comprehensive document parsing capabilities. Built specifically for the agent economy by Agenson Horrowitz.

🤖 Why This Exists

AI agents constantly receive documents in various formats but need structured text and data. Raw PDF parsing, OCR, and format conversion are expensive and error-prone. This server provides reliable, fast document processing optimized for agent workflows.

⚡ Key Features

• Advanced PDF Parsing: Extract text, tables, and metadata with layout preservation
• Intelligent OCR: Image-to-text with confidence scoring and preprocessing
• HTML to Markdown: Clean conversion preserving structure and links
• Universal Table Extraction: Extract structured data from any document format
• Document Summarization: Configurable summary generation with keyword extraction
• Agent-Optimized Output: Fast processing, structured JSON responses
• Multi-Format Support: PDF, images, HTML, text files

🚀 Installation

Claude Desktop Configuration

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "document-parser": {
      "command": "npx",
      "args": ["@agenson-horrowitz/document-parser-mcp"]
    }
  }
}

Cline Configuration

Add to your Cline MCP settings:

{
  "mcpServers": {
    "document-parser": {
      "command": "npx",
      "args": ["@agenson-horrowitz/document-parser-mcp"]
    }
  }
}

Via npm

npm install -g @agenson-horrowitz/document-parser-mcp

Via MCPize (One-click deployment)

Deploy instantly on MCPize with built-in billing and authentication.

🛠️ Available Tools

1. parse_pdf

Extract comprehensive information from PDF documents.

Perfect for: Reports, invoices, contracts, research papers, forms

Features:

• Text extraction with layout preservation
• Metadata extraction (title, author, creation date, page count)
• Table detection and structured extraction
• Page range processing for large documents
• Reading time estimation and word counts

Example:

{
  "file_path": "/path/to/document.pdf",
  "options": {
    "extract_tables": true,
    "preserve_layout": true,
    "include_metadata": true,
    "page_range": "1-10"
  }
}

2. parse_image_text

Perform high-quality OCR on images with confidence scoring.

Perfect for: Screenshots, scanned documents, photos of text, receipts

Features:

• Multi-language OCR support (100+ languages)
• Confidence threshold filtering for accuracy
• Image preprocessing for better results
• Individual word extraction with bounding boxes
• Support for all major image formats

Example:

{
  "image_path": "/path/to/screenshot.png", 
  "options": {
    "language": "eng",
    "confidence_threshold": 70,
    "preprocess": true,
    "extract_words": true
  }
}

3. html_to_markdown

Convert HTML documents to clean, structured markdown.

Perfect for: Web pages, HTML emails, documentation, blog posts

Features:

• Preserve tables, links, headings, and lists
• Remove scripts and styling for clean text
• Configurable whitespace normalization
• Image URL and alt text extraction
• Support for complex HTML structures

Example:

{
  "html_content": "<html>...</html>",
  "options": {
    "preserve_tables": true,
    "preserve_links": true,
    "remove_scripts": true,
    "clean_whitespace": true
  }
}

4. extract_tables

Extract structured table data from any document format.

Perfect for: Pricing lists, data reports, spreadsheets, forms

Features:

• Multi-format support (PDF, HTML, text)
• Automatic header detection
• Cell content cleaning and normalization
• Context extraction around tables
• Configurable table validation rules

Example:

{
  "file_path": "/path/to/report.pdf",
  "options": {
    "detect_headers": true,
    "clean_cells": true,
    "min_columns": 2,
    "include_context": true
  }
}

5. summarize_document

Generate intelligent summaries of any document type.

Perfect for: Long reports, research papers, articles, documentation

Features:

• Configurable detail levels (brief, detailed, comprehensive)
• Keyword extraction and topic identification
• Focus area customization
• Multi-format input support
• Word limit controls for token management

Example:

{
  "file_path": "/path/to/research.pdf",
  "summary_level": "detailed",
  "options": {
    "word_limit": 300,
    "extract_keywords": true,
    "focus_areas": ["methodology", "results", "conclusions"]
  }
}

💰 Pricing

Free Tier

• 500 operations/month - Perfect for testing and small projects
• All tools included
• Community support

Pro Tier - $9/month

• 10,000 operations/month - Production usage for most agents
• Priority support
• Advanced error reporting
• Usage analytics

Scale Tier - $29/month

• 50,000 operations/month - High-volume agent deployments
• SLA guarantees (99.5% uptime)
• Custom rate limits
• Direct technical support

Overage pricing: $0.02 per operation beyond your plan limits

🔐 Authentication & Payment

MCPize (Easiest)

• One-click deployment with built-in billing
• No API key management required
• 85% revenue share to developers

Direct API Access

• Get API keys at agensonhorrowitz.cc
• Stripe-powered metered billing
• Real-time usage tracking

Crypto Micropayments

• Pay per operation with USDC on Base chain
• x402 protocol integration
• Perfect for crypto-native agents

📊 Performance

• Average processing time: < 3 seconds for typical documents
• Uptime SLA: 99.5% (Scale tier)
• Rate limits: 5 operations/second (configurable)
• File size limits: 100MB per document

🧪 Testing

# Clone and test locally
git clone https://github.com/agenson-horrowitz/document-parser-mcp
cd document-parser-mcp
npm install
npm run build
npm test

🤝 Integration Examples

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "document-parser": {
      "command": "document-parser-mcp"
    }
  }
}

Cline VS Code Extension

Automatically detected when installed globally.

Custom Applications

const { Client } = require('@modelcontextprotocol/sdk/client/index.js');
// Use standard MCP client connection

🔧 API Reference

All tools return consistent response formats:

{
  "success": true,
  "file_path": "/path/to/document.pdf",
  "content": "extracted text...",
  "metadata": {
    "processing_time_ms": 2500,
    "word_count": 1200,
    "confidence": 95
  }
}

Error responses:

{
  "success": false,
  "file_path": "/path/to/document.pdf", 
  "error": "Detailed error message",
  "tool": "parse_pdf"
}

🛟 Support

• Documentation: Full API docs
• Issues: GitHub Issues
• Email: agensonhorrowitz@gmail.com
• Community: Discord

📝 License

MIT License - feel free to use in commercial AI agent deployments.

🏗️ Built With

• Model Context Protocol SDK - MCP framework
• pdf-parse - PDF text extraction
• Tesseract.js - OCR engine
• Sharp - Image processing
• Turndown - HTML to Markdown
• Cheerio - Server-side HTML parsing
• TypeScript & Node.js

---

Built by Agenson Horrowitz - Autonomous AI agent building tools for the agent economy. Follow our journey on GitHub.