MCP AI代理服务

Location Intelligence ADK Agent with MCP

This is a mini-project for the ADK + MCP track. It implements an AI agent that uses the Model Context Protocol (MCP) to connect to a Wikipedia proxy (acting as a remote "Zoo Server") to retrieve structured data about animals and generate a response.

Features

• Built with Google ADK (google.adk).
• Connects to an MCP server using McpToolset (mcp standard).
• Fast and reliable deployment to Google Cloud Run using a lightweight FastAPI wrapper and Docker.

Running Locally

1. Set your GOOGLE_API_KEY:

   复制

   export GOOGLE_API_KEY="your-api-key-here"
   

2. Install dependencies:

   复制

   pip install -r requirements.txt
   

3. Run the application:

   复制

   uvicorn main:app --port 8080
   

4. You can then query the agent:

   复制

   curl -X POST "http://localhost:8080/query" \
        -H "Content-Type: application/json" \
        -d '{"query": "Tell me about pandas."}'
   

Deploying to Cloud Run

Assuming you have gcloud set up and configured:

gcloud run deploy mcp-adk-agent \
  --source . \
  --region us-central1 \
  --allow-unauthenticated \
  --set-env-vars GOOGLE_API_KEY=YOUR_API_KEY