Building Advanced AI Agents with Strands: MCP, Hooks & Memory (Part 2)

In this advanced course, Dewan Lightfoot (AWS Developer Advocate) dives deep into the Strands Agents Python SDK. You will learn how to master the agentic loop, configure various model providers (AWS Bedrock, Anthropic, Ollama), and utilize the Model Context Protocol (MCP).

This tutorial moves beyond the basics, covering how to inject custom logic using Hooks, create self-extending agents that write their own tools, manage complex conversation sessions, and implement long-term persistent memory using Mem0.

What You Will Learn:
Agentic Loop: Understanding how Strands manages reasoning, planning, and execution.
Model Flexibility: Configuring agents with AWS Bedrock, Anthropic, and local models via Ollama.
Hooks: Injecting logic for logging and modifying agent behavior during lifecycle events.
Tools & MCP: Using built-in tools, creating custom tools, and integrating MCP servers (AWS Docs/Pricing).
State Management: Managing conversation history and sessions locally and via S3.
Long-Term Memory: Building memory-persistent agents using Mem0 and vector stores.

Prerequisites:
Python 3.10+ installed.
Access to an LLM (API key or Local/Ollama).

Timestamps:
0:00 - Introduction & Prerequisites
1:28 - Lab 1: Overview of Strands Agents & The Agentic Loop
14:14 - Lab 2: Model Providers (Bedrock, Anthropic, Ollama) & Configuration
26:11 - Lab 3: Advanced Response Processing with Hooks
39:37 - Lab 4: Custom Tools, MCP Integration & Self-Extending Agents
58:31 - Lab 5: Conversation, Session, and State Management
1:09:55 - Lab 6: Building Memory Persistent Agents with Mem0
1:25:01 - Summary & What's Next in Part 3