Building Smarter AI Agents: Memory Management with AgentCore

Agents can’t work without memory. Some memory might seem free inside of the agentic loop, but real systems need to manage memory over time. In this video, we’re going to explore agentic memory, short-term memory, long-term memory, and how to manage both in some practical code. Imagine walking into your favorite cafe. You say hello to the barista, order a cup of tea, and the barista picks up a mug, and then stares at you. Hello. They’ve forgotten what they’re doing, and they’ve forgotten what they’re going to do next. Super frustrating for both of you. And agents have the same needs.

Chapters

• 0:00 - Introduction to Agent Memory
• 2:30 - Short-Term Memory Explained
• 5:15 - Long-Term Memory Concepts
• 8:45 - Demo: Simple Agent with Strands SDK
• 12:30 - Integrating Amazon Bedrock Agent Core Memory
• 18:00 - Implementing Long-Term Memory Strategies
• 23:45 - Final Demo: Advanced Agent with Persistent Memory
• 28:30 - Conclusion and Key Takeaways

Links & Resources

• Full Code Example on Github