​
A2A-MCP Agents Workshop

In this workshop, you’ll build AI agents with xpander.ai and Amazon Bedrock. You’ll create agents that work together to help with coding and development tasks.

When finished, you’ll have a GitHub agent in your IDE, two coding agents, and a manager agent that coordinates them. You’ll use these agents to build and submit a sample application.

​
The xpander.ai Agentic Platform

xpander.ai is open-source multi-framework infrastructure for autonomous agents. It provides:

  • Unified state management (memory object) to make agents stateful, with thread support across frameworks
  • Out-of-the-box Agent-to-Agent protocol with orchestration rules between agents.
  • Tool access for AI models (Model Context Protocol (MCP) compatible)
  • Agentic RAG to optimize API responses.
  • Reliable execution engine for tool use and AI calls.
  • Dedicated AI workers for hosting custom agents of any framework or AI Model.
  • Human interfaces out-of-the-box - REST API, Web UI, MCP, and webhook support included.
  • Vault for keys.

​
Workshop Modules

  1. GitHub Agent - Build a GitHub search agent and connect it to your IDE through MCP.

  2. First Coding Agent - Create a developer agent with Amazon Bedrock that manages repos and writes code.

  3. Second Coding Agent - Add another developer agent with different capabilities and behavior.

  4. Manager Agent - Build a coordinator that assigns tasks and manages GitHub pull requests.

  5. Final Project - Use your agents to build and submit an application.

​
What You’ll Learn

This workshop teaches you:

  • When to use A2A for agent communication vs. MCP for tool access
  • How to use Agentic RAG to handle large API responses
  • Converting agents from different frameworks to work together
  • Adding tools to agents and setting up specific behaviors
  • Monitoring agent systems during operation

​
Modules

1. Supercharge Your IDE with MCP

Build a GitHub agent that connects to your IDE.

2. Build Your First Coding Agent

Create a Bedrock agent that handles code tasks.

3. Expand Your Agent Fleet

Add a second agent with different capabilities.

4. Build a Manager Agent

Create an agent that coordinates your other agents.

5. Build & Submit Your Application

Build and submit a project with your agents.

​
Prerequisites

You need:

  • AWS account with Amazon Bedrock access
  • xpander.ai account (provided at the workshop)
  • Cursor IDE with MCP support
  • Basic Python knowledge
  • GitHub account

Start with Module 1.