MCP Server with Google ADK Project

This project implements a Model Context Protocol (MCP) server using Google’s Agent Development Kit (ADK) for building intelligent agents and tools. The system features multiple specialized agents that collaborate through a coordinator agent to handle different types of requests.

Details in Medium Post

Project Structure

├── agents/                    # Contains agent implementations
│   ├── coordinator_agent.py   # LLM-based intelligent request router
│   ├── data_analysis_agent.py # Agent for analyzing data files with visualization
│   ├── search_agent.py        # Agent for web searches and information retrieval
│   └── code_generator_agent.py # Agent for generating code based on descriptions
├── tools/                     # Contains tool implementations
│   ├── code_generator_tool.py # Generates code in various programming languages
│   ├── data_analysis_tool.py  # Analyzes data and creates visualizations
│   ├── data_reader_tool.py    # Reads data from various file formats
│   ├── report_generator_tool.py # Generates formatted reports
│   └── web_search_tool.py     # Performs web searches
├── input_data/                # Directory for input data files (CSV, Excel)
├── analysis_output/plots/     # Generated data visualizations
├── reports/                   # Generated analysis reports
├── generated_code/            # Generated code outputs
├── server.py                  # MCP Server implementation using FastMCP
├── requests_log.txt           # Log of requests and responses
└── README.md                  # This file

Prerequisites

• Python 3.9+
• Groq and OpenRouter Key

Working Flow Architecture

Setup

1. Clone this repository
2. Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Set up your Groq & Openrouter Key: add key in .env if available otherwise create file as

touch .env 

add key

OPENROUTER_API_KEY="ADD_Your_Key"
GROQ_API_KEY="ADD_Your_Key" 

Running the MCP Server

python server.py

This will start the MCP server on http://0.0.0.0:8080

Available Agents

Coordinator Agent

The coordinator agent uses LLM-based routing to direct requests to the most appropriate specialized agent. It analyzes the content of the request and determines which agent can best handle it.

Search Agent

Performs web searches and provides information on various topics.

Example Request:

curl -X POST http://localhost:8080/ask -H "Content-Type: application/json" -d '{"query":"What is Model Context Protocol?"}'

Data Analysis Agent