AlloyDB MCP ADK Example

This example demonstrates how to interact with an AlloyDB database using AI Agents with the Model Context Protocol (MCP) Database Toolbox and Google Agent Development Kit (ADK).

Prerequisites

• Access to an AlloyDB instance on Google Cloud Platform
• Gemini API key for accessing Google AI models
• Python 3.10 or newer
• uv package manager

Setup

1. Synchronize dependencies using uv:

   uv sync

2. Create the environment files:

   cp .env.example .env
   cp alloydb_agent/.env.example alloydb_agent/.env

3. Edit the .env files and fill in your credentials:

   • In the root .env file:
     • Fill in AlloyDB connection details (ALLOYDB_PROJECT, ALLOYDB_REGION, etc.)
   • In the alloydb_agent/.env file:
     • For Google AI API (default):
       • Set GOOGLE_API_KEY with your Gemini API key
     • OR for Vertex AI:
       • See the ADK documentation for complete Vertex AI setup instructions

4. Source the environment variables:

   source .env

5. Download the MCP Toolbox for Databases

   export OS="linux/amd64" # one of linux/amd64, darwin/arm64, darwin/amd64, or windows/amd64
   curl -O https://storage.googleapis.com/genai-toolbox/v0.6.0/$OS/toolbox
   chmod +x toolbox

6. Start the MCP Toolbox with the provided tools configuration:

   ./toolbox --tools-file "alloy-tools.yaml"

7. In a separate terminal, activate the virtual environment and start the ADK server:

   source .venv/bin/activate
   adk web

   The ADK web interface should now be available at http://localhost:8000, where you can interact with your AlloyDB database using natural language.

Disclaimer

This is example/demo code and may not be actively maintained or ready for production deployment. Use at your own risk.