Simple MCP Server

🚀 Simple MCP Information Server

This is a microservice based on FastMCP, offering search tools for Wikipedia and Arxiv, implemented using LangChain community utilities.

This project demonstrates how to use FastMCP to create an MCP (Model Context Protocol) server. It provides two simple tools for Claude Desktop to search Wikipedia and retrieve academic paper information from Arxiv.

🚀 What is MCP?

Model Control Protocol (MCP) is a protocol that allows language models (such as Claude) to communicate with external tools (referred to as "capabilities") in a standard way. These capabilities can be:

• Functions written in any programming language
• Web APIs
• Local scripts

Through MCP, Claude can call your tools, obtain responses, and continue the conversation.

✨ Features

This information server exposes two tools via MCP:

1. get_info(searchterm: str) — Search for a short summary on Wikipedia
2. get_research_paper(searchterm: str) — Retrieve academic paper metadata from Arxiv

⚡ Overview

This tool exposes two compatible endpoints via MCP:

• get_info: Search for information on Wikipedia.
• get_research_paper: Retrieve academic paper information from Arxiv.

🔧 Technical Details

• FastMCP: Used to create the MCP server.
• LangChain Community Toolkit: Used to interact with Wikipedia and Arxiv.
• Python 3.12+: The base language for running the project.

📦 Installation

Prerequisites

• Python 3.12 or higher
• The uv tool installed and configured (for running the MCP server)
• Claude Desktop installed (for connection and testing)

Initialize the Project

1. Create a new project:

   uv init simple_mcp_server
   cd simple_mcp_server
   

2. Activate the virtual environment:

   source env/bin/activate  # For Mac/Linux
   # Or
   .\env\Scripts\activate  # For Windows
   

3. Install dependencies:

   pip install mcp[cli] langchain-community wikipedia arxiv
   

💻 Usage Examples

Basic Usage

Search Wikipedia

from langchain_community.wikipedia import WikipediaAPI

# Initialize the Wikipedia API tool
wiki_tool = WikipediaAPI()

# Call the search function
result = wiki_tool.search("Donald Trump")
print(result)

Get Academic Papers

import arxiv

# Initialize the Arxiv client
client = arxiv.Client()

# Search for papers
papers = client.search("AI in finance", max_results=5)
for paper in papers:
    print(paper.title, paper.authors)

Advanced Usage

Use a Custom MCP Client

import os
from langchain_community import wikipedia, arxiv

# Initialize the Wikipedia and Arxiv tools
wiki_tool = WikipediaAPI()
arxiv_tool = arxiv.Client()

# Create an MCP agent
class MCPAgent:
    def __init__(self):
        self.tools = [wiki_tool, arxiv_tool]

    async def run(self, query):
        # Select the appropriate tool based on the query
        if "Wikipedia" in query:
            return wiki_tool.search(query)
        else:
            return arxiv_tool.search(query)

# Initialize the agent
agent = MCPAgent()

# Execute the search task
result = await agent.run("tell me about donald trump")
print(f"\nResult: {result}")

📚 Configuration

Edit the configuration file of Claude Desktop and add the following content:

"display": ":1"

Start Claude Desktop and connect to the MCP server.

🎯 Advantages

• No Dependency on Claude Desktop UI: Use command-line tools directly.
• Easy to Replace LLM: Supports multiple language models, such as Groq, OpenAI, etc.
• Full Control of Interaction Flow: Customize agent logic and tool call order.

📝 Conclusion

This project demonstrates how to quickly create a powerful information retrieval server using FastMCP. Whether for personal research or enterprise applications, it can be flexibly extended and customized according to requirements.