Engineering Production AI Agents with Amazon Bedrock AgentCore
This two-day, instructor-led course, offered in person or virtually, teaches experienced builders how to develop and evaluate an agent prototype for production use on Amazon Bedrock AgentCore.
Description
This two-day, instructor-led course, offered in person or virtually, teaches experienced builders how to develop and evaluate an agent prototype for production use on Amazon Bedrock AgentCore. Participants choose between the AgentCore harness and AgentCore Runtime for a given workload. Across both days, they build an agent on Runtime, connect it to tools through Gateway, and give it short-term and long-term memory. They authenticate users and outbound calls with Identity, constrain tool calls with Policy, and measure and improve the agent with Observability, Evaluations, and optimization. The capstone joins several agents over the Agent-to-Agent (A2A) protocol and publishes them to AWS Agent Registry.
Who This Course Is For
- Software engineers and ML engineers who have built an agent prototype and must run it in production
- Solutions architects designing agent platforms on AWS
- Platform and DevOps engineers who will operate, secure, and monitor agents
- Technical leads choosing between managed agent options on AWS
What You Will Learn
- Select and deploy an agent hosting model on Amazon Bedrock AgentCore that fits a workload's control, isolation, and session requirements.
- Connect agents to tools and context through AgentCore Gateway, the built-in tools, and AgentCore Memory.
- Secure and govern agent access with AgentCore Identity and AgentCore Policy.
- Operate, measure, and improve agents and multi-agent systems in production with AgentCore Observability, Evaluations, optimization, and AWS Agent Registry.
Course Outline
- Choosing an Agent Platform on AWS
- Hosting Agents on AgentCore Runtime
- Connecting Tools with AgentCore Gateway
- Giving Agents Memory
- Securing Agents with AgentCore Identity
- Governing Tool Calls with AgentCore Policy
- Observing, Evaluating, and Improving Agents
- Capstone — A Governed Multi-Agent System
This is a high-level overview. For the complete syllabus with detailed topics and lab descriptions,request the full syllabus.
Prerequisites
- LF-20303 _AI Agent Fundamentals on AWS_, or equivalent experience building a tool-using agent
- Working Python: functions, packages, virtual environments, async basics
- Comfort with the AWS CLI, IAM roles and policies, AWS Lambda, and Amazon CloudWatch
- Familiarity with OAuth 2.0 and JSON Web Tokens (JWT) at the level of reading a token and a scope
- A modern web browser and a reliable internet connection; the lab sandbox is provided
Delivery Options
- Live, instructor-led
- Virtual instructor-led
Bring This Course to Your Team
This course is delivered as private, instructor-led training for teams and organizations. Contact us for a quote, scheduling, and group options.