ArXiv Search MCP Server

This is a Model Context Protocol (MCP) server designed to search for papers on arXiv.

Features

• Provides an MCP tool to search arXiv based on keywords.
• Retrieves paper details including title, authors, abstract, published date, and PDF link.
• Formats the results in Markdown.

Provided Tool: search_arxiv

This server exposes a single tool named search_arxiv.

Input Schema

{
  "type": "object",
  "properties": {
    "keywords": {
      "type": "string",
      "description": "Keywords to search for (e.g., \"machine learning\")"
    },
    "max_results": {
      "type": "number",
      "description": "Maximum number of papers to retrieve (default: 10)",
      "minimum": 1
    }
  },
  "required": [
    "keywords"
  ]
}

• keywords (string, required): The search terms to use for finding papers.
• max_results (number, optional): The maximum number of papers to return. Defaults to 10 if not specified.

Output

The tool returns a Markdown formatted string containing the search results.

Usage with Cline

Once this server is configured in your MCP settings, you can ask Cline to use it.

Examples:

• "Search arXiv for 'quantum computing' papers"
• "Find 5 papers on 'large language models' on arXiv"
• "Use the arxiv-search tool with keywords 'transformer architecture' and max_results 3"

Cline will interact with this MCP server via the search_arxiv tool to fetch and display the requested papers.

Development

• Language: TypeScript
• Core Dependency: arxiv-api for interacting with the arXiv API.
• Build: Compiled to JavaScript using npm run build.

Running Locally

This server is intended to be run by an MCP client (like the one integrated into the VS Code extension). The client manages starting and stopping the server process based on the configuration in cline_mcp_settings.json.

To build the server manually:

npm install
npm run build

The executable will be located at build/index.js.

License

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