home / mcp / agentic ai mcp server

Agentic AI MCP Server

AgenticAI using MCP

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "dev484p-agenticai_mcp": {
      "command": "uv",
      "args": [
        "run",
        "mcp",
        "dev",
        "server.py"
      ],
      "env": {
        "GROQ_API_KEY": "YOUR_GROQ_API_KEY",
        "TAVILY_API_KEY": "YOUR_TAVILY_API_KEY"
      }
    }
  }
}

You can run an MCP-powered agent that combines a Groq-hosted LLM with tools for Wikipedia data, live web results via Tavily, and Yahoo Finance data using a standardized Model Context Protocol server. This setup lets the AI fetch contextual information on demand and perform complex data-backed tasks through a clean client–server workflow.

How to use

Start the MCP server and the client, then interact with the AI to perform searches, retrieve current stock data, and browse factual information. The server exposes three integrated tools via MCP: a Wikipedia search tool for factual references, an Internet search tool powered by Tavily for broad web results, and a Yahoo Finance tool for real-time stock and financial data. You interact with the client to issue queries; the LLM fetches context from the tools through the MCP server and returns informed responses.

How to install

Prerequisites you need to set up first are installing UV tooling, obtaining a Groq API key, and obtaining a Tavily API key.

Follow these concrete steps to install and run the MCP server and client.

git clone https://github.com/dev484p/AgenticAI_MCP
cd AgenticAI_MCP
uv add "mcp[cli]"

# Set up your API keys in a configuration file (example structure shown below)
# Open keys.json and add your keys as shown in the example
{
  "groqApiKey": "YOUR_GROQ_API_KEY",
  "tavilyApiKey": "YOUR_TAVILY_API_KEY"
}

Security and usage notes

This setup relies on host-mediated authentication and secure transport for MCP interactions. Keep your API keys private, rotate keys periodically, and restrict access to the MCP endpoints to trusted clients.

If you want to debug or develop locally, you can run the development inspector alongside the MCP server to monitor tool discovery and request routing.

Available tools

wiki_search

Query Wikipedia for factual information and summaries

internet_search

Perform web searches via Tavily to retrieve comprehensive results and sources

finance_yahoo

Fetch real-time stock prices and financial data from Yahoo Finance