Managing Short-Term and Long-Term Memory With Amazon Bedrock AgentCore

Without memory, an AI agent treats each question independently. If you ask it to find flights from Seattle to Tokyo and then follow up with, "what about business class?", it will not know what you mean. Amazon Bedrock AgentCore Memory addresses this in two ways: short-term memory preserves context within a session, while long-term memory extracts durable facts and preferences for use across sessions.

In this lab, you will build an Amazon Bedrock AgentCore harness with memory enabled and put both mechanisms to the test. You will attach a pre-configured gateway that gives the agent a flight search tool, hold a multi-turn conversation that only works if the agent remembers, and then verify that a stated travel preference survives into a brand new session. Finally, you will inspect the raw memory events and extracted memory records in CloudWatch Logs.

Learning objectives

Upon completion of this intermediate-level lab, you will be able to:

• Configure an Amazon Bedrock AgentCore harness with memory enabled
• Test short-term and long-term memory behavior in the harness playground
• Inspect memory events and extracted memory records

Intended audience

• Candidates for the AWS Certified Generative AI Developer - Professional certification
• Cloud Architects
• Software Engineers

Prerequisites

Familiarity with the following will be beneficial but is not required:

• Amazon Bedrock AgentCore

The following content can be used to fulfill the prerequisites:

• Streamlining Workflows and Tools With Amazon Bedrock AgentCore