ADK MCP Agent Backend

This mini-project demonstrates how to implement a native Model Context Protocol (MCP) integration with Google's ADK (Agent Development Kit using google-generativeai).

It features a responsive, premium web interface and an autonomous gemini-2.5-flash agent that dynamically executes MCP tools to retrieve structured JSON records from a local simulated HR Database container.

Project Structure

• app.py: FastAPI server containerized for Google Cloud Run deployment.
• mcp_server.py: The FastMCP standard IO backend pretending to act as an external database.
• mcp_client.py: Bridges rigid asynchronous MCP protocols smoothly into the standard environment.
• agent.py: Uses google-generativeai ADK to bind the isolated MCP environment as an immediate tool for Gemini logic routing.

Setup & Running

1. Initialize dependencies:

   pip install -r requirements.txt
   

2. Create .env and add:

   GOOGLE_API_KEY="YOUR_GEMINI_KEY"
   

3. Run Local Server:

   python app.py
   

   Visit http://localhost:8080 to experience the dynamic styling layout and interact with the AI.

Deploying to Cloud Run

Execute inside a configured Google Cloud environment:

gcloud builds submit --tag gcr.io/YOUR_PROJECT_ID/adk-mcp-agent
gcloud run deploy adk-mcp-agent \
  --image gcr.io/YOUR_PROJECT_ID/adk-mcp-agent \
  --platform managed \
  --allow-unauthenticated \
  --set-env-vars=GOOGLE_API_KEY="YOUR_KEY"