MCP Paper Analysis System

Model Context Protocol (MCP) implementation for academic paper research and analysis using the Anthropic Claude API and ArXiv integration.

Overview

This on-going project demonstrates the implementation of the Model Context Protocol (MCP) to create an intelligent paper analysis system that can:

• Search for academic papers on ArXiv
• Extract and analyze paper content
• Generate comprehensive summaries
• Provide comparative topic analysis
• Save analysis results to files

Features

🔍 Paper Search & Management

• Search ArXiv database by topic
• Automatically organize papers by topic
• Store paper metadata (title, authors, abstract, publication date)
• Generate unique paper IDs for easy reference

📊 Advanced Analysis

• Comprehensive Paper Summaries: Extract key information including:

  • Abstract and problem statement
  • Methodology and frameworks used
  • Existing solutions and their limitations
  • Proposed improvements and innovations
  • Experimental results and findings
  • Key contributions and limitations

• Topic-Based Analysis: Comparative analysis across multiple papers including:

  • Common themes and research trends
  • Methodological approaches comparison
  • Research gaps identification
  • Practical applications assessment

Data Persistence

• Save summaries and analyses to formatted text files
• Organized file structure with timestamps
• Support for both basic and comprehensive summary formats

Interactive Chat Interface

• Natural language queries
• Tool-based response system
• Real-time paper analysis
• Conversational interface for research assistance

Project Structure

claude_mcp/
├── mcp_tutorial.ipynb          # Main notebook with MCP implementation
├── papers/                     # Paper storage directory
│   ├── machine_learning/       # Topic-based subdirectories
│   │   └── papers_info.json   # Paper metadata
│   └── [other_topics]/
├── summaries/                  # Generated analysis files
│   ├── [paper_id]_comprehensive_summary.txt
│   ├── [paper_id]_basic_info.txt
│   └── [topic]_topic_analysis.txt
└── README.md                   # This file

Installation

1. Clone the repository:

   git clone [repository-url]
   cd claude_mcp

2. Install required packages:

   pip install arxiv anthropic python-dotenv PyPDF2 requests

3. Set up API credentials:

   • Create a .env file in the project root
   • Add your Anthropic API key:

     ANTHROPIC_API_KEY=your_api_key_here
     

   • Or directly edit the ANTHROPIC_API_KEY variable in the notebook NOT ADVISABLE

Usage

Getting Started

1. Open the Jupyter notebook:

   jupyter notebook mcp_tutorial.ipynb

2. Execute the setup cells to import libraries and initialize the system

3. Start the chat interface by running the chat_loop() function

Available Tools

The system provides six main tools accessible through natural language queries:

Tool	Description	Example Query
search_papers	Search ArXiv for papers on a topic	"Search for papers about machine learning"
extract_info	Get basic paper information	"Get info for paper 1909.03550v1"
summarize_paper	Generate comprehensive analysis	"Summarize paper 1909.03550v1"
analyze_papers_by_topic	Compare papers in a topic	"Analyze machine learning papers"
save_summary_to_file	Save paper analysis to file	"Save summary of paper XYZ to file"
save_topic_analysis_to_file	Save topic analysis to file	"Save machine learning analysis to file"

Example Queries

Query: Search for papers about deep learning
Query: What are the key findings in paper 2106.09685v2?
Query: Analyze all papers on natural language processing
Query: Save comprehensive summary of paper 1909.03550v1
Query: Compare methodologies in computer vision papers

Dependencies

• arxiv: ArXiv API integration
• anthropic: Claude API client
• PyPDF2: PDF text extraction
• requests: HTTP requests for PDF download
• python-dotenv: Environment variable management
• json: Data serialization
• os: File system operations