About This Book
Agentic coding is what happens when an AI system stops completing your next line and starts executing a plan: reading your codebase, running commands, testing its own output, and coming back with a result instead of a suggestion. This book teaches that shift from the ground up: what an agent actually is, how to talk to one, how to break a task down into something small enough for it to handle, and how to check its work before you trust it.
Who This Book Is For
You're learning to program and want to build with AI agents from day one, instead of learning "the old way" first and bolting agents on later. Or you already know how to code but haven't worked with an agent that plans and executes multi-step tasks on its own, and want real grounding in how that changes the job.
Either way, if you want to get past typing prompts into a chat window and into an actual working practice, planning tasks, reviewing an agent's output, choosing tools, building small multi-agent systems, this book is written for you. You should be comfortable reading code, even if you're still learning to write much of it yourself. Nothing here assumes you've used an AI coding tool before.
What You’ll Learn
You’ll explore the foundations and practice of agentic coding through examples, walkthroughs, and conceptual discussions:
This book teaches you to work inside that shift, from someone who has never touched an AI coding tool to someone who can plan a task, hand it to an agent, and know exactly what to check before trusting what comes back.
You build real, runnable projects: a document summarizer and a retrieval-augmented FAQ agent against the Anthropic and OpenAI SDKs directly, then a small multi-agent pipeline where a planner, a writer, and a reviewer each run as a separate model call. Along the way you work with Claude Code, OpenAI's Codex CLI, GitHub Copilot, Cursor, and Pi, a minimal open-source agent short enough to read start to finish.
The book is honest about what changes fast and what doesn't. Specific tools, prices, and benchmark standings sit in dated callouts. What's built to last is the practice: decomposing a task, writing a requirements contract, reviewing an agent's output critically, and catching its mistakes before they ship, skills that carry over to whatever tool you're using two years from now.
Part I builds the mental model. Part II turns it into a practice, planning, tool use, memory, and your first working agents. Part III covers the patterns production teams actually run, including a chapter that tells you plainly when you're reading ahead of where you need to be yet.
Remember, Agentic coding doesn't replace all your engineering tasks, treat it as a assistive tool to make your life easy.
Why It Belongs on Leanpub
Leanpub’s iterative publishing model aligns perfectly with the spirit of agentic development. Every specific tool, price, and benchmark number in this book sits in a dated callout for a reason: this space moves faster than any print run. Leanpub's update model is what makes that honesty practical instead of theoretical, buy once, and future updates catch you up on what actually changed instead of leaving you with a book that quietly went stale.
How to Read This Book
Read it front to back if you're starting from zero. Jump to the chapter you need if you already have the basics down.
Open your editor, pick a real task, and put what's in this book to work on it. That's the only way any of it becomes a habit instead of a summary you read once.
