Agentic Programming Patterns: Building Intelligent AI Systems

A comprehensive guide to designing, implementing, and deploying AI agents using proven architectural patterns.

About This Book

This book provides a deep dive into the design patterns and architectural principles behind modern AI agentic systems. As large language models (LLMs) have become increasingly capable, the field has moved beyond simple prompt-and-response interactions toward sophisticated agent architectures that can plan, reason, use tools, collaborate, and iteratively refine their outputs.

Drawing from the latest research and industry practices from Anthropic, OpenAI, DeepLearning.AI, and the open-source community, this book catalogs the essential patterns that power today’s most effective AI agents — from simple prompt chaining to complex multi-agent orchestration systems.

Who This Book Is For

• Software engineers looking to integrate LLM-powered agents into their applications
• AI/ML engineers wanting to move beyond simple chat interfaces to agentic architectures
• Technical architects designing production-ready AI agent systems
• Researchers seeking a practical overview of agentic design patterns
• Engineering managers who need to understand the landscape of agent architectures to make informed technical decisions

What You’ll Learn

• The foundational building blocks of agentic systems (augmented LLMs, tools, memory)
• Core agentic design patterns: Reflection, Tool Use, Planning, and Multi-Agent Collaboration
• Workflow patterns: Prompt Chaining, Routing, Parallelization, Orchestrator-Workers, and Evaluator-Optimizer
• How to manage agent memory, context windows, and state across interactions
• Practical implementation using popular frameworks (LangChain, CrewAI, AutoGen)
• How to design effective Agent-Computer Interfaces (ACI) and prompt-engineer your tools
• Production best practices: guardrails, safety, observability, cost control, and human-in-the-loop oversight
• Real-world case studies in coding agents, customer support, research assistants, and data pipelines

Book Structure

The book progresses from foundational concepts to advanced patterns and production considerations:

• Chapters 0–2 introduce agentic programming, the augmented LLM building block, and the agent loop
• Chapters 3–6 cover the four core design patterns: Reflection, Tool Use, Planning, and Multi-Agent Collaboration
• Chapters 7–11 present workflow-level patterns: Prompt Chaining, Routing, Parallelization, Orchestrator-Workers, and Evaluator-Optimizer
• Chapters 12–13 address agent memory/state and tool design (ACI)
• Chapter 14 surveys the major frameworks and SDKs
• Chapter 15 covers production deployment, safety, and governance
• Chapter 16 presents real-world case studies and composite architectures

Author Notes

The agentic AI space is evolving extremely rapidly. The patterns described in this book represent the state of the art as of early 2026, synthesized from Anthropic’s “Building Effective Agents” guide, Andrew Ng’s Agentic Design Patterns series, framework documentation from LangChain, CrewAI, and AutoGen, and current academic research. While specific APIs and frameworks will continue to evolve, the underlying architectural patterns are likely to remain relevant as foundational building blocks.

All code examples use Python unless otherwise specified.

Table of Contents

• Chapter 0 – Introduction to Agentic Programming
• Chapter 1 – The Augmented LLM
• Chapter 2 – The Agent Loop
• Chapter 3 – Reflection Pattern
• Chapter 4 – Tool Use Pattern
• Chapter 5 – Planning Pattern
• Chapter 6 – Multi-Agent Collaboration
• Chapter 7 – Prompt Chaining
• Chapter 8 – Routing Pattern
• Chapter 9 – Parallelization Pattern
• Chapter 10 – Orchestrator-Workers Pattern
• Chapter 11 – Evaluator-Optimizer Pattern
• Chapter 12 – Agent Memory and State
• Chapter 13 – Designing Agent-Computer Interfaces
• Chapter 14 – Frameworks and SDKs
• Chapter 15 – Production, Safety, and Governance
• Chapter 16 – Case Studies and Composite Architectures