MaksPyn/arxiv-mcp-server

arXiv MCP Server

An enhanced Model Context Protocol (MCP) server that provides comprehensive access to arXiv for searching, downloading, reading, and analyzing scientific papers with AI-powered research prompts.

Overview

This MCP server enables AI assistants to not only search and retrieve scientific papers from arXiv but also download them locally, extract content, and generate research-oriented analysis prompts. Perfect for researchers, students, and anyone working with academic literature.

Features

Core Features

• Search papers by title, author, abstract, category, or any combination
• Retrieve papers by their arXiv ID
• Get recent papers in specific categories
• Find all papers by a specific author
• Rate limiting compliance with arXiv API guidelines

Enhanced Features (v1.1.0)

• 📥 Paper Download & Storage: Download and store papers locally with metadata tracking
• 📖 PDF Content Reading: Extract and read text content from downloaded papers
• 🔍 In-Paper Search: Search for specific terms within downloaded papers
• 🤖 AI Research Prompts: Pre-configured analysis prompts for paper summarization, methodology analysis, and more
• 📊 Storage Management: Track storage usage and manage downloaded papers
• 🔧 Enhanced Error Handling: Robust error handling with retry logic
• 📝 Structured Logging: Winston-based logging with configurable levels

Installation

Prerequisites

• Node.js (v16 or higher)
• npm or yarn

Option 1: Install from npm

npm install -g @arxiv/mcp-server

Then run:

arxiv-mcp-server

Option 2: Install from Source

1. Clone this repository:

git clone https://github.com/MaksPyn/arxiv-mcp-server.git
cd arxiv-mcp-server

2. Install dependencies:

npm install

3. Build the TypeScript code:

npm run build

Option 3: Docker

1. Using Docker Compose:

cd docker
docker-compose up -d

2. Using Docker directly:

docker build -t arxiv-mcp-server -f docker/Dockerfile .
docker run -it --name arxiv-mcp-server arxiv-mcp-server

Configuration

Copy the example environment file and customize as needed:

cp config/.env.example .env

Available configuration options:

• LOG_LEVEL: Logging level (debug, info, warn, error)
• STORAGE_DIR: Directory for storing downloaded papers
• MAX_STORAGE_SIZE_MB: Maximum storage size for papers
• RATE_LIMIT_DELAY_MS: Delay between API requests (default: 3000ms)

Usage

Running the Server

Start the server using:

npm start

Or for development (build + run):

npm run dev

Integrating with Claude Desktop

Add the following to your Claude Desktop configuration file:

Windows: %APPDATA%\Claude\claude_desktop_config.json macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Linux: ~/.config/claude/claude_desktop_config.json

{
  "mcpServers": {
    "arxiv": {
      "command": "node",
      "args": ["C:/path/to/arxiv-server/build/index.js"]
    }
  }
}

Replace C:/path/to/arxiv-server with the actual path to your arxiv-server directory.

Available Tools

1. search_papers

Search for papers using various criteria.

Parameters:

• query (string, optional): Raw search query
• title (string, optional): Search in paper titles
• author (string, optional): Search by author name
• abstract (string, optional): Search in abstracts
• category (string, optional): Filter by category (e.g., "cs.AI", "math.CO")
• all (string, optional): Search in all fields
• start (number, optional): Starting index (default: 0)
• maxResults (number, optional): Maximum results to return, max 2000 (default: 10)
• sortBy (string, optional): Sort by "relevance", "lastUpdatedDate", or "submittedDate" (default: "relevance")
• sortOrder (string, optional): "ascending" or "descending" (default: "descending")

Example:

{
  "title": "neural networks",
  "category": "cs.AI",
  "maxResults": 5
}

2. get_paper_by_id

Retrieve specific papers by their arXiv IDs.

Parameters:

• ids (array of strings, required): Array of arXiv IDs

Example:

{
  "ids": ["2301.00001", "2312.12345"]
}

3. get_recent_papers

Get the most recent papers in a specific category.

Parameters:

• category (string, required): Category to filter by (e.g., "cs.AI", "physics.quant-ph")
• maxResults (number, optional): Maximum results (default: 10)

Example:

{
  "category": "cs.LG",
  "maxResults": 20
}

4. search_author

Find all papers by a specific author.

Parameters:

• author (string, required): Author name to search for
• maxResults (number, optional): Maximum results (default: 20)
• sortBy (string, optional): "submittedDate" or "lastUpdatedDate" (default: "submittedDate")

Example:

{
  "author": "Yann LeCun",
  "maxResults": 10
}

5. download_paper

Download a paper PDF and store it locally.

Parameters:

• arxivId (string, required): arXiv ID of the paper to download

Example:

{
  "arxivId": "2301.00001"
}

6. list_downloaded_papers

List all locally downloaded papers with metadata.

Parameters: None

7. delete_paper

Delete a downloaded paper from local storage.

Parameters:

• arxivId (string, required): arXiv ID of the paper to delete

8. get_storage_stats

Get storage statistics for downloaded papers.

Parameters: None

Returns:

• Total number of papers
• Total storage size
• Formatted size string

9. read_paper_content

Read and extract text content from a downloaded paper.

Parameters:

• arxivId (string, required): arXiv ID of the paper to read

Returns:

• Extracted sections (abstract, introduction, methodology, results, discussion, conclusion, references)
• Full text length

10. search_in_paper

Search for text within a downloaded paper.

Parameters:

• arxivId (string, required): arXiv ID of the paper to search in
• searchTerm (string, required): Text to search for
• caseSensitive (boolean, optional): Whether the search should be case sensitive (default: false)

Returns:

• Total matches found
• Matches organized by section with context

11. get_analysis_prompts

Get available research analysis prompts.

Parameters: None

Returns:

• List of available prompts with IDs, names, descriptions, and required variables

12. analyze_paper

Generate an analysis prompt for a paper.

Parameters:

• arxivId (string, required): arXiv ID of the paper to analyze
• promptId (string, required): ID of the analysis prompt to use

Available Prompt IDs:

• summary: Comprehensive paper summary
• key_findings: Extract key findings and results
• methodology_analysis: Analyze research methodology
• literature_review: Create literature review entry
• research_gaps: Identify research gaps and future directions
• technical_deep_dive: Technical analysis with algorithms and implementation details
• comparison: Compare with related work
• practical_applications: Identify real-world applications

Example:

{
  "arxivId": "2301.00001",
  "promptId": "summary"
}

arXiv Categories

Common arXiv category codes include:

Computer Science

• cs.AI - Artificial Intelligence
• cs.LG - Machine Learning
• cs.CV - Computer Vision and Pattern Recognition
• cs.CL - Computation and Language
• cs.NE - Neural and Evolutionary Computing

Physics

• physics.gen-ph - General Physics
• physics.optics - Optics
• quant-ph - Quantum Physics

Mathematics

• math.CO - Combinatorics
• math.PR - Probability
• math.ST - Statistics Theory

For a complete list, visit arXiv Category Taxonomy.

Response Format

All tools return structured JSON responses containing:

• Paper IDs
• Titles
• Authors (with affiliations when available)
• Publication/update dates
• Categories
• Summaries (truncated for search results, full for specific paper retrieval)
• Direct PDF links
• Additional metadata (DOI, journal references, comments)

Rate Limiting

This server respects arXiv's API usage guidelines by implementing a 3-second delay between requests. This is handled automatically and ensures compliance with arXiv's terms of service.

Error Handling

The server includes comprehensive error handling for:

• Network errors
• Invalid parameters
• Rate limiting
• arXiv API errors

Errors are returned in a structured format that can be easily parsed by the MCP client.

Development

Project Structure

arxiv-server/
├── src/
│   ├── index.ts              # Main server implementation
│   ├── types/                # TypeScript type definitions
│   ├── utils/                # Utility functions (logger, errors, file operations)
│   ├── storage/              # Storage management for downloaded papers
│   └── prompts/              # Research prompt management
├── build/                    # Compiled JavaScript files
├── storage/                  # Downloaded papers storage
├── prompts/                  # Prompt templates
├── logs/                     # Application logs
├── config/                   # Configuration files
├── docker/                   # Docker configuration
│   ├── Dockerfile
│   └── docker-compose.yml
├── package.json              # Project dependencies
├── tsconfig.json             # TypeScript configuration
├── .npmignore                # NPM publish ignore rules
└── README.md                 # This file

Building from Source

npm run build

Running in Development Mode

npm run dev

License

MIT License - see package.json for details.

Contributing

Contributions are welcome! Please feel free to submit issues or pull requests.

Acknowledgments

This server uses the arXiv API to access scientific papers. Please respect arXiv's terms of service and usage guidelines when using this server.