Mark Scott UX

Lab 2: Build a Local Prototype with Strands SDK

Create a sophisticated mortgage assistant using the Strands SDK. This prototype demonstrates multi-agent orchestration with specialized sub-agents working together to handle different aspects of mortgage inquiries.

What You’ll Build

• Supervisor Agent — orchestrates tasks and routes requests to appropriate specialists
• New Application Agent — handles mortgage application processes and guidance
• General Mortgage Agent — answers questions using your knowledge base
• Existing Mortgage Agent — manages account inquiries and updates

Lab Commands

source
cd 02-Strands-mortgage-assistant/
uv run mortgage_agent.py --prompt

What You’ll Learn

• Multi-Agent Orchestration — coordinating multiple specialized agents
• Agents-as-Tools Pattern — using agents as composable tools
• Knowledge Base Integration — connecting agents to Bedrock Knowledge Base
• Local Development — building and testing agents in your dev environment

Lab 3: AgentCore Runtime

Take your local mortgage assistant prototype and deploy it to Amazon Bedrock AgentCore Runtime — a secure, serverless, and purpose-built hosting environment for production AI agents with automatic scaling and framework flexibility.

What You’ll Build

• Serverless Deployment — no infrastructure management required
• Secure Hosting — enterprise-grade security and isolation
• Scalable Execution — automatic scaling based on demand
• Framework Flexibility — support for Strands, LangGraph, and other frameworks

Notebook

03-agentcore-runtime/agentcore_runtime.ipynb

What You’ll Learn

• Production Deployment — moving from prototype to production
• Containerization — packaging agents for cloud deployment
• Runtime Configuration — setting up AgentCore Runtime
• Scalability Patterns — building agents that scale automatically