Day 8: Agentic AI - Bedrock Agents + AgentCore
Learning Objectives
- - Distinguish Bedrock Agents (managed) from AgentCore (deployment platform)
- - Know all 9 AgentCore services and their exam triggers
- - Understand MCP server implementation patterns (Lambda vs ECS)
- - Design multi-agent orchestration (supervisor vs routing mode)
- - Implement Strands Agents SDK with tool use
Tasks
Tasks
0/6 completed- Read45m
Amazon Bedrock AgentCore Developer Guide
All 9 AgentCore services: Runtime, Gateway, Policy, Identity, Memory, Observability, Evaluations, Code Interpreter, Browser.
- Read30m
Bedrock Agents User Guide
Managed agent service. Action groups, tool use, knowledge base integration.
- Blog25m
Strands Agents SDK Technical Deep Dive
Architecture, tool definitions, observability for the Strands framework.
- Blog25m
Multi-Agent Orchestration with Reasoning
Supervisor mode vs routing mode. Key for multi-agent architecture questions.
- Blog20m
MCP on AWS - Unlocking Model Context Protocol
AWS API MCP Server, IAM permissions, CloudTrail auditing for MCP.
- Watch30m
Mike Chambers AgentCore Full Walkthrough
Hands-on deploy, Gateway, Memory, Observability demo by AWS Dev Advocate.
Exam Skills
Write your understanding, then reveal the reference answer.
Hands-On Lab
Build real muscle memory with these activities.
Build a Bedrock Agent with Action Groups
Create a Bedrock Agent with a Lambda-backed action group that can perform a real-world task.
- 1 Open Bedrock console → Agents → Create agent
- 2 Set the agent instructions: 'You are a helpful assistant that can look up order status and process returns'
- 3 Create a Lambda function that simulates order lookup (returns mock data)
- 4 Define an action group with an OpenAPI schema describing the getOrderStatus and processReturn actions
- 5 Test the agent in the console: 'What is the status of order 12345?' and verify it calls the Lambda function
Explore Bedrock Agents Quickstart Workshop
Work through the quickstart workshop to build an agent for Well-Architected queries.
- 1 Clone the amazon-bedrock-agents-quickstart repository
- 2 Follow the setup instructions to deploy the Lambda function and agent
- 3 Test the agent with Well-Architected queries
- 4 Examine how the action group schema maps to Lambda invocations
- 5 Review the agent trace to understand the ReAct reasoning loop
Scenarios
Think through each scenario before revealing the answer.
DevOps Monitoring Agent Architecture
- •Does this need a managed agent (Bedrock Agents) or custom framework?
- •How do you expose existing Python scripts as agent tools?
- •How does the agent learn from past incident remediations?
Financial Analysis Agent with Code Interpreter
- •Which AgentCore service provides sandboxed code execution?
- •How do you securely access internal APIs?
- •What libraries are available in the sandbox?
MCP Server for CRM Integration
- •What is the difference between stateless and stateful MCP servers?
- •Which compute service is best for stateless vs stateful?
- •How do you make MCP servers available to agents?
Practice Questions
5 questions across 3 difficulty levels.
Further Reading
Go deeper into today's topics.
InterWorks — Bedrock Agents Walkthrough
Coffee shop agent with tool use, step-by-step console screenshots.
Bedrock Agents Action Groups
Lambda-backed tools for real-world tasks with detailed walkthrough.
Multi-Agent Collaboration — GA Announcement
Supervisor-collaborator architecture: designate supervisor agent, associate collaborators for complex workflows.
Introducing Strands Agents — Open Source SDK
Model-driven agent framework: build agents in a few lines of code with @tool decorators and streaming.
MCP vs Strands vs RAG vs A2A — Agents Compared
Comprehensive comparison of all agent options on AWS.