arxiv-search-mcp

arxiv-search-mcp

3
research_and_data

The arXiv Search MCP Server is designed to allow users to search and retrieve academic papers from the arXiv.org repository. It provides functionalities such as category-based search and fetching the latest papers, enabling efficient research access.

arXiv Search MCP Server

An MCP server that provides tools to search and fetch papers from arXiv.org.

Features

  • Search papers by category
  • Get latest papers sorted by submission date
  • Formatted output with title, authors, summary, and link

Development

Prerequisites

  • Deno installed on your system
  • MCP compatible environment

Setup

  1. Clone the repository
  2. Install dependencies:
deno cache --reload src/main.ts

Running the Server

Development mode with file watching:

deno task dev

Build executable:

deno task compile

Integration with Claude Desktop

Add the following configuration to your claude_desktop_config.json:

{
  "mcpServers": {
    "arxiv-search-mcp": {
      "command": "/path/to/dir/arxiv-search-mcp/bin/arxiv-search-mcp"
    }
  }
}

Replace /path/to/dir with the actual path to your compiled binary.

Usage

Example usage screenshot: Sample usage with Claude

The server provides a tool named search_arxiv that accepts the following parameters:

{
  "category": string,    // arXiv category (e.g., cs.AI, cs.LG, astro-ph)
  "max_results": number  // Number of papers to fetch (1-100, default: 5)
}

Example

Request:

{
  "category": "cs.AI",
  "max_results": 5
}

This will return the 5 most recent papers from the Artificial Intelligence category.

Available Categories

Some popular arXiv categories:

  • cs.AI: Artificial Intelligence
  • cs.LG: Machine Learning
  • cs.CL: Computation and Language
  • cs.CV: Computer Vision
  • cs.NE: Neural and Evolutionary Computing
  • cs.RO: Robotics
  • astro-ph: Astrophysics
  • physics: Physics
  • math: Mathematics
  • q-bio: Quantitative Biology

For a complete list of categories, visit arXiv taxonomy.

License

This project is licensed under the MIT License - see the file for details.