Prajapdh/arxiv_mcp_server

Academic Paper Search with MCP

This project demonstrates the Model Context Protocol (MCP) architecture with a server that provides academic paper search capabilities and a client that connects to multiple MCP servers in a chatbot interface.

Project Overview

This application consists of:

1. MCP Server (ResearchServer.py): Provides tools, resources, and prompts for searching and exploring academic papers from arXiv.
2. MCP Client (MCPChatbotWithMultipleServers.py): A chatbot interface that can connect to multiple MCP servers and use their tools, resources, and prompts.

Features

MCP Server

• 🔍 Search Papers: Search for academic papers on arXiv by topic
• 📋 Extract Info: Get detailed information about specific papers
• 📂 Browse Resources: View paper topics and contents through URI-based resources
• 📝 Prompts: Generate structured prompts for research tasks

MCP Client

• 🤖 Chatbot Interface: Natural language interface to the MCP servers
• 🌐 Multi-Server Support: Connect to multiple MCP servers simultaneously
• 🧰 Resource Navigation: Browse paper topics with @ commands
• ⚡ Prompt Execution: Run predefined prompts with / commands

Prerequisites

• Python 3.9+
• Node.js (for the MCP Inspector)

Installation

1. Clone this repository:

git clone https://github.com/Prajapdh/arxiv_mcp_server.git
cd arxiv_mcp_server

2. Install dependencies:

pip install -r requirements.txt

3. Create a .env file with your Anthropic API key:

ANTHROPIC_API_KEY=your_api_key_here

4. Configure your servers in server_config.json:

{
  "mcpServers": {
    "research": {
      "command": "uv",
      "args": ["run", "ResearchServer.py"],
      "env": null
    }
  }
}

Running the MCP Server

Run with stdio transport (default)

uv run ResearchServer.py

Run with MCP Inspector for debugging

npx @modelcontextprotocol/inspector uv run ResearchServer.py

Running the MCP Client

uv run MCPChatbotWithMultipleServers.py

Chatbot Commands

• Regular query: Type any text to chat with the assistant
• @folders: List all available paper topic folders
• @<topic>: Browse papers in a specific topic
• /prompts: List all available prompts
• /prompt <name> <arg1=value1>: Execute a specific prompt with arguments
• exit: Exit the chatbot

Server Features

Tools

Resources

Prompts

Project Structure

DeepLearning MCP tutorial/
├── MCPChatbotWithMultipleServers.py  # Multi-server MCP client
├── ResearchServer.py                 # ArXiv paper search MCP server
├── server_config.json                # Server connection configuration
├── requirements.txt                  # Project dependencies
├── papers/                           # Directory for storing paper information
│   └── {topic}/                      # Topic-specific directories
│       └── papers_info.json          # Stored paper metadata
└── README.md                         # This file

Understanding MCP

The Model Context Protocol (MCP) is a standard for connecting large language models (LLMs) to external tools and resources. In this project:

• The MCP server exposes tools (functions), resources (data sources), and prompts (templates)
• The MCP client connects to these servers and enables the LLM to use their capabilities
• Communication happens via either stdio (for local use) or SSE (for network use)

Extending the Project

• Add more MCP servers with different capabilities
• Implement additional paper analysis tools
• Add support for other academic databases
• Create visualization tools for research data

Troubleshooting

• If you encounter connection errors, ensure the server is running before starting the client
• For SSE transport issues, check port availability or switch to stdio transport
• Verify your Anthropic API key is correctly set in the .env file

License