Inspiration
The Online Boutique demo is great, but it’s just a static microservices app.
We wanted to supercharge it with agentic AI, helping users discover eco-friendly products while showing how AI can extend microservices without touching the core code.
What it does
- Fetches products from the Boutique catalog via MCP
- Filters for sustainability using eco-tags and carbon scores
- Sends shortlisted products to the Recommender Agent via A2A for final ranking
- Returns a green, ranked recommendation list to the user
- Provides a bookmarklet widget for instant integration on Online Boutique
How we built it
- Base app: Online Boutique (GCP microservices demo)
- Platform: Google Kubernetes Engine (GKE)
- Agents: SustainableAdvisorAgent + RecommenderAgent using ADK
- Protocols: MCP for catalog access, A2A for agent communication
- Containerized: Each agent runs as its own microservice
- Frontend: Lightweight JavaScript widget with glassmorphism design
Challenges we ran into
- Service-to-service communication errors in Kubernetes DNS
- JSON parsing and API response formatting issues from the Catalog Service
- Managing multiple containerized agents and ensuring proper MCP/A2A integration
- Keeping everything demo-ready under tight time constraints
Accomplishments that we're proud of
- Built a true agentic AI microservice on top of an existing system
- Demonstrated inter-agent communication and sustainability ranking
- Successfully deployed multiple AI agents in containers on GKE
What we learned
- How to extend microservices externally without touching core code
- Deploying AI-powered agents at scale using GKE and ADK
- Handling MCP and A2A protocols in real-world scenarios
- Designing lightweight, cross-platform UI widgets for instant integration
What's next for Sustainable Shopping Advisor
- Gemini AI integration for richer natural language understanding
- Mobile SDK and voice assistant integration for shopping on the go
- Adding social sustainability metrics and carbon footprint calculators
Built With
- a2a
- adk
- gke
- mcp
- python

Log in or sign up for Devpost to join the conversation.