What you'll learn

AI agents act autonomously, and scoping permissions for which actions they take can be quite complex. Not only do you have to consider the agent’s bounds and capabilities, but often you must also consider the permissions (data access, actions) of the principle invoking the agent. In this course, Agentic Solutions Patterns for AWS, you’ll gain the ability to integrate autonomous agents into modern AWS architecture while maintaining the principle of least privilege. First, you’ll explore the inherent risks of agent autonomy, and the means by which constraints can be applied. Next, you’ll discover how to resolve hosting, orchestration, memory, identity, and tool brokering into Amazon Bedrock AgentCore's Harness, Runtime, Memory, Identity, and Gateway capabilities. Finally, you’ll learn how to enforce permission and audit guarantees outside the model with AgentCore Policy and Observability, and view the resulting design through the AWS Well-Architected Agentic AI Lens. When you’re finished with this course, you’ll have the skills and knowledge of agentic AI engineering needed to build secure, capable, integrated, and scalable agents on AWS.