mcp-a2a-codelabs
This repository contains implementations and artifacts for three Google Cloud AI Agents Development Kit (ADK) codelabs. Each project demonstrates how to build and deploy intelligent agents using MCP (Model Context Protocol) and A2A (Agent-to-Agent) communication.
The goal of this assignment is to reproduce all three codelabs, deploy the systems, and document the working implementations.
📂 Repository Structure
- codelab1-multiagent-adk-a2a — Multi-agent image scoring system built with ADK and A2A
- codelab2-currency-agent — Currency conversion agent integrating MCP, ADK, and A2A
- codelab3-purchasing-concierge-a2a — Purchasing concierge coordinating multiple seller agents via A2A
- README.md — Consolidated documentation for all projects
🧩 Overview of Codelabs
🧠 Codelab 1 — Create Multi-Agent System with ADK and A2A
Codelab link: https://codelabs.developers.google.com/codelabs/create-multi-agents-adk-a2a#0
Focus: Building an image scoring multi-agent system using ADK, deploying it to Vertex AI Agent Engine, and testing inter-agent communication through A2A.
Agents include:
- Image generation agent
- Scoring agent
- Checker agent
Artifacts include packaged models, deployment logs, and test-client results.
💱 Codelab 2 — Getting Started with ADK, MCP, and A2A
Codelab link: https://codelabs.developers.google.com/codelabs/currency-agent#0
Focus: Integrating Model Context Protocol (MCP) with ADK. Implements a currency exchange agent that communicates with a local MCP server to retrieve exchange rates and exposes an A2A endpoint for inter-agent communication.
Deliverables:
- MCP rate server
- Currency agent with A2A adapter
- Test results of currency conversions
🍔 Codelab 3 — Getting Started with A2A Action Engine (Purchasing Concierge)
Codelab link: https://codelabs.developers.google.com/intro-a2a-purchasing-concierge#0
Focus: Multi-agent orchestration using the A2A Action Engine. Includes the burger-agent and pizza-agent deployed on Cloud Run, coordinated by a purchasing concierge agent that runs on Vertex AI Agent Engine.
Demonstrates complete multi-agent collaboration using A2A for ordering and fulfillment workflows.
🚀 Environment and Deployment Summary
All three codelabs use:
- Google Cloud Project:
specialtopics018221801 - Region:
us-central1 - Vertex AI Agent Engine for managed agent deployment
- Cloud Run for remote seller agents
- MCP server for external data integration
- A2A protocol for agent-to-agent messaging
🎥 Video Demonstration
YouTube Playlist: Youtube Link
Each video in the playlist demonstrates:
- Environment setup and configuration
- Agent deployment and runtime testing
- MCP and A2A interactions
- Verified successful outputs
📁 Submission Artifacts
| Directory | Description |
|---|---|
codelab1-multiagent-adk-a2a/ | Image scoring multi-agent ADK project |
codelab2-currency-agent/ | MCP server and currency exchange agent |
codelab3-purchasing-concierge-a2a/ | Cloud Run seller agents and A2A concierge |
README.md | Main documentation |
| YouTube Playlist | Recorded video walkthroughs |
✅ Deliverables Checklist
- [x] Working MCP server and A2A agents
- [x] Multi-agent deployment to Vertex AI Agent Engine
- [x] Cloud Run services for burger and pizza agents
- [x] Organized GitHub repo with all logs and configurations
- [x] YouTube playlist demonstrating execution
🧾 References
- AI Agents Development Kit (ADK)
- Model Context Protocol (MCP)
- Agent-to-Agent (A2A) Specification
- Vertex AI Agent Engine Documentation
👤 Author
Apurva Karne Graduate Student — Special Topics 018221801 Google Cloud ADK / MCP / A2A Multi-Agent Systems Assignment