📚 Simple MCP Info Server

A FastMCP-powered microservice that provides search tools for Wikipedia and Arxiv using the LangChain Community utilities.

---

This project demonstrates how to create and run an MCP (Model Context Protocol) server using FastMCP. It provides simple tools that Claude Desktop can use to search Wikipedia and fetch academic papers from Arxiv.

---

🚀 What is MCP?

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

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

Using MCP, Claude can invoke your tools, get the response, and continue the conversation.

---

📚 What This Project Does

This Info Server exposes two tools via MCP:

1. get_info(searchterm: str) — Searches Wikipedia for short summaries
2. get_research_paper(searchterm: str) — Searches Arxiv for academic paper metadata

---

⚡ Overview

This tool exposes two FastMCP-compatible endpoints:

• get_info: Fetches summarized information from Wikipedia.
• get_research_paper: Retrieves academic paper details from Arxiv.

Designed for plug-and-play use in modular AI systems or agent runtimes.

---

🧰 Setup Guide

✅ Prerequisites

• Python 3.12+
• uv installed

  Install via curl -LsSf https://astral.sh/uv/install.sh | sh or brew install astral-sh/uv/uv

  Windows powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

• Claude Desktop

  'Claude Desktop' installed

---

🛠️ Project Initialization

uv init simple_mcp_server
cd simple_mcp_server
uv venv
source .venv/bin/activate

📦 Install Dependencies

uv add "mcp[cli]"
uv add langchain_community
uv add wikipedia
uv add arxiv

server.py (FastMCP) Runs your Wikipedia + Arxiv search server locally.

🚀 Run the Server Ensure you’re in the virtual environment, then:

uv run server.py

The server will run over stdio, ready to be called as an MCP tool.Just verify if there are any error.

Option 1: Update claude_desktop_config.json (for Claude Desktop)

Run this command to open the config file in VS Code (or use any editor):

code ~/Library/Application\ Support/Claude/claude_desktop_config.json

Replace or add the following mcpServers block:

{
  "mcpServers": {
    "info-server": {
      "command": "/Users/yuvrajfirodiya/.local/bin/uv",
      "args": [
        "--directory",
        "/Users/yuvrajfirodiya/Source/Python-Langchain-Projects/simple_mcp_server",
        "run",
        "server.py"
      ]
    }
  }
}

✅Verify the following:

command: Make sure it points to your uv binary (which uv in terminal to confirm).

args: Points to the root of your MCP project.

"run", "server.py": You’re running the correct server file.

Or

Option 2: Use a Custom MCP Client (Host)

You can also run the server using a custom MCP client built with the mcpclient library.

This is useful for advanced workflows, testing, or when integrating with your own LLM setup.

📄 client.py includes the full server specification and config Set up the client using a structured config method like this: import asyncio import os from dotenv import load_dotenv from langchain_groq import ChatGroq from mcp_use import MCPAgent, MCPClient

import asyncio
import os
from dotenv import load_dotenv
from langchain_groq import ChatGroq
from mcp_use import MCPAgent, MCPClient

async def main():
    # Load environment variables (e.g., GROQ_API_KEY)
    load_dotenv()

    # Define the MCP server config
    config = {
        "mcpServers": {
            "info-server": {
                "command": "/Users/yuvrajfirodiya/.local/bin/uv",
                "args": [
                    "--directory",
                    "/Users/yuvrajfirodiya/Source/Python-Langchain-Projects/simple_mcp_server",
                    "run",
                    "server.py"
                ],
                "env": {
                    "DISPLAY": ":1"
                }
            }
        }
    }

    # Initialize the client
    client = MCPClient.from_dict(config)

    # Initialize the Groq LLM (Llama 3)
    llm = ChatGroq(
        model_name="Llama3-8b-8192",
        streaming=True
    )

    # Build your agent
    agent = MCPAgent(llm=llm, client=client, max_steps=30)

    # Ask a question
    result = await agent.run("tell me about donald trump")
    print(f"\nResult: {result}")

if __name__ == "__main__":
    asyncio.run(main())

✅ Benefits of This Approach 🔓 No dependency on Claude Desktop UI

🧩 Easily swap out LLMs (Groq, OpenAI, etc.)

🧠 Full control over how your agent interacts with tools

⚙️ Configurable, scriptable, and scalable