An MCP server that provides access to arXiv papers through their API.

This server allows LLM clients (like Claude Desktop) to:

• Search for scientific papers on arXiv by title and abstract content
• Control search result sorting (by submission date, update date, or relevance)
• See total matching results count to gauge search specificity
• Get paper metadata and abstracts
• Retrieve full paper text as Markdown (converted from PDF)
• Access links to available paper formats (PDF/HTML)
• Browse and search arXiv subject categories

To use the web version just add this connector to Claude.ai https://mcp.andybrandt.net/arxiv . You can also install & use it locally.

This server can be run in two modes: as a local stdio server for desktop clients or as a network-accessible web server.

To install Simple Arxiv for Claude Desktop automatically via Smithery:

npx -y @smithery/cli install mcp-simple-arxiv --client claude

pip install mcp-simple-arxiv

Add this configuration to your claude_desktop_config.json:

(Mac OS)

{
  "mcpServers": {
    "simple-arxiv": {
      "command": "python",
      "args": ["-m", "mcp_simple_arxiv"]
      }
  }
}

(Windows version):

{
  "mcpServers": {
    "simple-arxiv": {
      "command": "C:\\Users\\YOUR_USERNAME\\AppData\\Local\\Programs\\Python\\Python311\\python.exe",
      "args": [
        "-m",
        "mcp_simple_arxiv"
      ]
    }
  }
}

After restarting Claude Desktop, the following capabilities will be available:

You can ask Claude to search for papers using queries like:

Can you search arXiv for recent papers about large language models?

The LLM can sort results by submission date, update date, or relevance:

Search arXiv for papers about transformers, sorted by relevance.

Search results include:

• Total number of matching papers (e.g., "Found 15,234 total results, showing first 10")
• Paper title, authors, and arXiv ID
• Categories (primary and additional)
• Publication date
• Abstract preview

Once you have a paper ID, you can ask for more details:

Can you show me the details for paper 2103.08220?

This will return:

• Full paper title
• Authors
• Publication and update dates
• Journal reference (if available)
• Paper abstract
• Links to available formats (PDF/HTML)

For web deployment see WEB_DEPLOYMENT.md.

To install for development:

git clone https://github.com/andybrandt/mcp-simple-arxiv
cd mcp-simple-arxiv
pip install -e .

This server follows arXiv API usage guidelines:

• Rate limiting to max 1 request per 3 seconds
• Single connection at a time
• Proper error handling and retry logic

MIT