Strands SDK Agent - Learning Repository

A hands-on learning repository for building production-ready AI agents using AWS Bedrock, Strands SDK, and Amazon Bedrock AgentCore.

Overview

This repository contains tutorials and examples demonstrating how to build, deploy, and monitor agentic AI applications using:

• Amazon Bedrock - Managed foundation models
• Strands SDK - Python framework for building AI agents
• Bedrock AgentCore - Serverless runtime for production deployment

Repository Structure

├── ecommerce_agent/        # Production e-commerce support agent
│   ├── main.py            # Agent implementation with tools, memory, and guardrails
│   └── requirements.txt   # Python dependencies
├── images/                # Architecture diagrams and screenshots
├── sessions/              # Agent session data (gitignored)
├── 1_introduction.ipynb   # GenAI & AWS Bedrock basics
├── 2_strands_core_concepts.ipynb  # Agent loops, state, sessions
├── 3_tools.ipynb          # Python tools, MCP, executors
├── 3.1-lambda-agentcore-gateway.ipynb  # Gateway integration
├── 4_multi_agent_patterns.ipynb  # Multi-agent orchestration
├── 5_observability.ipynb  # Monitoring and telemetry
├── 6_gaurdrail.ipynb      # Content filtering and safety
└── 7_deployment.ipynb     # Production deployment

E-commerce Agent Features

The sample agent (ecommerce_agent/main.py) demonstrates:

• Guardrails - Content filtering and PII protection
• Memory - AgentCore Memory for conversation context
• Tools - MCP gateway integration + custom Python tools
• Observability - OpenTelemetry integration with Langfuse
• OAuth - M2M authentication for gateway access

Prerequisites

• AWS Account with Bedrock access
• Python 3.10+
• AWS CLI configured
• IAM permissions for Bedrock runtime

Quick Start

# Install dependencies
pip install strands-agents bedrock-agentcore boto3

# Run notebooks
jupyter notebook 1_introduction.ipynb

# Deploy agent
cd ecommerce_agent
python main.py

Learning Path

1. Introduction - GenAI fundamentals and Bedrock basics
2. Core Concepts - Agent loops, state management, sessions
3. Tools - Integrate Python functions and MCP servers
4. Multi-Agent - Swarm, workflow, and graph patterns
5. Observability - Traces, metrics, and monitoring
6. Guardrails - Safety and compliance controls
7. Deployment - Production deployment with AgentCore

Key Concepts

Agentic AI - Autonomous systems that perceive, reason, act, and learn to achieve goals without constant human oversight.

Strands SDK - Framework providing agent loops, tool orchestration, memory management, and multi-agent patterns.

AgentCore - Serverless runtime offering managed execution, memory, identity, and observability for production agents.

Resources

• AWS Bedrock Documentation
• Strands SDK GitHub
• AgentCore Documentation