AWS GenAI Builder Series - AWS AgentCore Services

AWS GenAI Builder Series — AWS AgentCore Services

Format: 2 hours (presentation → live demo)
Focus: Build and operate AI agents on AWS using Amazon Bedrock AgentCore. We’ll work directly from the official AgentCore Samples repo to take an agent from local run to a production-ready deployment pattern.

What you’ll learn

• Where AgentCore fits in an AWS stack: Runtime, Gateway (tooling), Memory, Identity, and Observability
• How to run an agent locally, then deploy it on AWS
• How to register and call tools (Lambda/HTTP) through the Gateway
• How to add state with Memory and secure access with Identity
• Practical ops: tracing, debugging, guardrails, and cost awareness

Live demo (using the GitHub samples)

1. Runtime quickstart: run a minimal agent locally → deploy → invoke
2. Gateway tool call: expose a simple Lambda/HTTP tool and call it from the agent
3. Add Memory + Observability: persist context and trace steps end-to-end
4. (If time) wire up Identity for a third-party call

Who should attend
Engineers and architects building agentic applications on AWS (SWE, MLE, SRE, platform).

Prerequisites (bring your laptop)

• AWS account with access to Bedrock and AgentCore
• AWS CLI configured (aws configure)
• Python 3.10+ and Docker/Finch for local dev
• Access to at least one Bedrock model (e.g., Claude)
• We’ll reference: awslabs/amazon-bedrock-agentcore-samples

Agenda (2 hours)

• 15 min — Context & goals
• 35 min — Theory walkthrough (architecture, patterns, gotchas)
• 50 min — Live demo (zero → working agent)
• 10 min — Q&A + pitfalls
• 10 min — Homework lab & repo pointers

Takeaways

• Slides and a reproducible demo repo
• A mini runbook for local → cloud deployment
• A checklist to productionize your first AgentCore agent