Advanced Strands Agents with MCP

A comprehensive advanced course for building production-ready AI agents using the Strands Agents SDK. This repository contains 6 progressive labs that teach advanced capabilities including tool integration, memory persistence, Model Context Protocol (MCP), and comprehensive observability.

🎯 Course Overview

This advanced course provides foundational and advanced expertise in building production-ready AI agents, focusing on the agent class, agentic loop, and the complete ecosystem of components that power intelligent autonomous systems. You'll master core principles of agentic AI, understanding how agents combine large language models, tools, and memory to create systems that can understand, plan, and execute actions autonomously.

Course Topics:

• Strands Agents SDK - Advanced agent architecture and lifecycle management
• Model Context Protocol (MCP) - Standardized tool and service integration
• Multi-Provider Configuration - Amazon Bedrock, Anthropic, OpenAI, and Ollama
• Advanced Processing - Hooks, session management, and conversation strategies
• Memory Systems - Long-term persistent memory with FAISS, OpenSearch, and Mem0
• Enterprise Features - Observability, metrics analysis, and performance optimization

Prerequisites: Basic understanding of AI/ML concepts. For beginners, we recommend starting with Getting Started with Strands Agents (Course 1).

📚 Lab Structure

Lab 1: Overview of Strands Agents (12:52)

Files: first_agent.py

Learn fundamental agentic AI concepts and build your first Strands agent:

• Basic agent creation with default configuration (no API keys required)
• Core agent components and execution flow
• Agent result examination (message, metrics, state, stop reasons)
• Dynamic model configuration and system prompt modification
• Conversation history management and message clearing

Lab 2: Model Providers and Configuration (11:59)

Files: anthropic_model.py, bedrock_model.py, ollama_model.py, openai_model.py

Configure agents across multiple LLM providers for flexibility and cost optimization:

• Model architecture overview and provider-specific parameters
• Bedrock model setup with structured output capabilities
• Anthropic model configuration with thinking mode
• Ollama local deployment and OpenAI integration
• Metrics analysis and performance monitoring

Lab 3: Advanced Response Processing with Hooks (13:30)

Files: async_example.py, hook_example_1.py, hook_example_2.py

Implement custom logic to intercept and modify agent behavior at lifecycle points:

• Event-driven hook system and lifecycle management
• Before/after event handling and agent modifications
• Async iterators, callback handlers, and retry logic
• Tool hook examples and precision parameter setup

Lab 4: Tools and MCP Integration (18:55)

Files: mcp_integration.py, self_extending_example.py, tools/

Extend agent capabilities with custom tools and external service integration:

• Built-in tools from strands-agents-tools library
• Custom tool creation using @tool decorator
• MCP server configuration for AWS Documentation and Pricing
• Self-extending agents and meta tooling capabilities
• Proper error handling and security implementation

Lab 5: Conversation and Session Management (11:26)

Files: session_example.py, verify_session.py

Manage conversation state and context effectively across interactions:

• Context window challenges and management strategies
• Three conversation manager approaches (Null, SlidingWindow, Summarizing)
• Session state persistence and user isolation
• File-based and Amazon S3 session storage options

Lab 6: Memory Persistent Agents (15:19)

Files: memory_example.py

Build agents with long-term memory capabilities across conversations:

• Memory backends integration (FAISS, OpenSearch, Mem0)
• Web search integration with DuckDuckGo
• Memory storage, retrieval, and relevance scoring
• Amazon Bedrock Knowledge Bases integration
• Retention policies and privacy controls

📖 Learning Path

1. Start with Lab 1 - Learn agent fundamentals with no setup required
2. Progress through Labs 2-3 - Configure multiple providers and implement hooks
3. Master Lab 4 - Integrate tools and MCP servers for external capabilities
4. Build with Lab 5 - Implement sophisticated conversation management
5. Advanced Lab 6 - Create agents with persistent memory systems

🔧 Model Provider Support

This course primarily uses the Anthropic Claude API, but all examples can be configured to work with:

• Amazon Bedrock (Claude, Llama, Titan, and other models)
• OpenAI (GPT-4, GPT-3.5-turbo)
• Ollama (Local model deployment)
• Other providers supported by Strands SDK

🚀 Quick Start

Prerequisites

• Python 3.8+
• Virtual environment (recommended)
• Anthropic Claude API key (primary requirement)
• Additional API keys for specific labs:
  • Amazon Bedrock (for AWS integration labs)
  • OpenAI (optional alternative)
  • Mem0 (for memory persistence Lab 6)

Getting Your Anthropic API Key

1. Visit Anthropic Console
2. Sign up for an account or log in
3. Navigate to API Keys section
4. Click Create Key and give it a name
5. Copy your API key (starts with sk-ant-)

Installation