The Encyclopedia of Agentic Coding Patterns is a compendium of tested solutions (“patterns”) for building software with AI agents. The entries run from the strategic (what to build, and why) to the tactical (how to direct an agent through the work). Whether you’ve never written a line of code or you’ve been shipping software for years, they meet you where you are. Each pattern is self-contained, so you can read them in any order and combine them to fit your situation.
Browse the Encyclopedia
Introduction — Get your bearings. This book is an encyclopedia, not a tutorial, and these pages explain how it is organized, who it is for, and how to read it. Includes Welcome, What Is Agentic Coding?, What Are Design Patterns?, How to Read This Book, Pattern Map, and more. View all 8 entries →
Product Judgment and What to Create — Decide what deserves to exist before writing any code. The strategic layer: what to build, who it is for, and why anyone will care. Includes Problem, Customer, Value Proposition, Beachhead, Product-Market Fit, Zero to One, and more. View all 19 entries →
Intent, Scope, and Decision-Making — Pin down what you are actually building and how you will decide among competing options. The patterns an agent needs before it can do useful work. Includes Brief, Requirement, Specification, Acceptance Criteria, Design Doc, Tradeoff, and more. View all 12 entries →
Structure and Decomposition — Give a system a skeleton. How to divide a program into parts that compose well, with clean seams that humans and agents can both reason about. Includes Architecture, Abstraction, Module, Interface, Boundary, Coupling, Cohesion, and more. View all 18 entries →
Data, State, and Truth — Every system remembers things. These patterns cover the shape of data, where state lives, and how to keep different parts of a system from disagreeing about what is true. Includes Data Model, State, Source of Truth, Schema, Transaction, Bounded Context, Ubiquitous Language, and more. View all 27 entries →
Computation and Interaction — Software computes and it communicates. These patterns describe how programs transform data and how separate pieces of software talk to each other. Includes Algorithm, API, Protocol, Determinism, Side Effect, Concurrency, Event, and more. View all 8 entries →
Correctness, Testing, and Evolution — Software changes constantly: new features, bug fixes, shifting requirements. These tactical patterns cover how you know your code is correct, how you keep it correct as it changes, and how you notice when it breaks. Includes Test, Invariant, Test-Driven Development, Refactor, Observability, Feedback Loop, Strangler Fig, and more. View all 37 entries →
Security and Trust — Not all actors are friendly, not all inputs are well-formed, not all code does what it claims. Security is behaving correctly under attack; trust is deciding what to rely on and what to verify. Includes Threat Model, Least Privilege, Prompt Injection, Sandbox, Blast Radius, RAG Poisoning, Agent Trap, and more. View all 19 entries →
Human-Facing Software — Every system eventually meets a person. Patterns for the moment software touches a human being: perception, cognition, communication, and access. Includes UX, Affordance, User Feedback, Accessibility, Internationalization, and more. View all 6 entries →
Operations and Change Management — Software that works on your laptop isn’t finished. These patterns govern how code moves into the world, how it evolves once it gets there, and how you roll back when something goes wrong. Includes Deployment, Continuous Integration, Continuous Delivery, Feature Flag, Rollback, Runbook, Cascade Failure, and more. View all 14 entries →
Socio-Technical Systems — Software is built by people, and the shape of the organization shows up in the shape of the code. Patterns for team structure, ownership, and the human layer of the system. Includes Conway’s Law, Team Cognitive Load, Ownership, Stream-Aligned Team, Platform as a Product, Inverse Conway Maneuver, and more. View all 10 entries →
Design Heuristics and Smells — Rules of thumb for decisions where the right answer depends on context, and the warning signs that tell you something has gone wrong. Taste and pattern recognition in one place. Includes KISS, YAGNI, Local Reasoning, Code Smell, AI Smell, Jagged Frontier, Vibe Coding, and more. View all 16 entries →
Agentic Software Construction — The newest layer of practice: building software with and through AI agents that read code, propose changes, run commands, and iterate under human guidance. The largest section in the book. Includes Model, Prompt, Context Engineering, Agent, Tool, MCP, Subagent, Skill, Orchestrator-Workers, and more. View all 51 entries →
Agent Governance and Feedback — Agents take actions on their own, sometimes good ones and sometimes not. Patterns for approval, evaluation, and the feedback loops that let you trust an agent over time. Includes Approval Policy, Eval, Human in the Loop, Bounded Autonomy, Steering Loop, Agent Sprawl, AgentOps, and more. View all 22 entries →