🧠 arXiv Research Assistant MCP Server

This project is an MCP (Model Context Protocol) server built to interact with the vast arXiv.org paper database.

It allows clients like Claude AI to search, explore, and compare arXiv papers efficiently — all through a custom-built, local server. It’s built with Python and the FastMCP framework, and uses uv for lightweight package management.

✨ Features

• 🔍 Keyword-based Paper Search
  Search arXiv papers by keywords, with options to sort by relevance or most recent.

• 📚 Latest Papers by Category
  Specify an arXiv category code (e.g., cs.AI, math.AP) to fetch the most recent papers in that field.

• 📄 Paper Details Lookup
  Fetch detailed metadata using a paper's arXiv ID: title, authors, abstract, categories, DOI, PDF link, and more.

• 🧑‍🔬 Author-based Paper Search
  Retrieve a list of papers published by a specific author.

• 📊 Trend Analysis (Experimental)
  Get an overview of trending keywords or topics based on recent papers in a category (currently uses mock data).

• 📝 Summarization Prompt Generator
  Dynamically generate prompts that help LLMs summarize a selected paper more effectively.

• 🆚 Comparison Prompt Generator
  Provide two paper IDs to generate a structured prompt for comparing their content.

🛠️ Tech Stack

• Python 3.11+
• FastMCP
• uv (for dependency & environment management)
• requests (for API communication)
• xml.etree.ElementTree (for parsing XML responses)

🚀 Getting Started

Installing via Smithery

To install arXiv Research Assistant MCP Server for Claude Desktop automatically via Smithery:

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

Installation from PyPI

uv pip install arxiv-paper-mcp

🔧 Clone the repository (for development)

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

🔧 Install Dependencies (for development)

Use uv to install all dependencies in editable mode:

uv pip install -e .

⚙️ How to Run

▶️ Run the server (locally)

arxiv-paper-mcp

🔌 Use with Claude

To use this MCP server with Claude, add the following JSON configuration to Claude's MCP settings:

{
  "mcpServers": {
    "arXivPaper": {
      "command": "uv",
      "args": [
        "tool",
        "run",
        "arxiv-paper-mcp"
      ]
    }
  }
}

Project Structure

arxiv-mcp-server/
├── arxiv_mcp/              # Main package
│   ├── __init__.py
│   ├── app.py              # FastMCP app setup
│   ├── server.py           # Server entry point
│   ├── utils.py            # arXiv API communication logic
│   ├── resources/          # MCP resources (categories, authors, etc.)
│   ├── tools/              # MCP tools (search, detail lookup, trends)
│   └── prompts/            # Prompt templates (summarize, compare)
├── pyproject.toml          # Project config & dependencies
└── README.md               # This file