A New Book: Claude Code Projects
I am thrilled to announce my new book, Claude Code Projects: Build 10 Real-World Projects with Skills, Hooks, MCP, Testing, Git, and Production Workflows, published by Packt Publishing and launching December 9, 2026.
This is not another “learn the prompts” book. It is a hands-on, project-by-project guide to treating Claude Code as a real member of your engineering team — one that reads your codebase, opens branches, runs tests, and opens pull requests you can actually review.
📘 Get Claude Code Projects on Amazon → · 📅 Releases December 9, 2026 · 🎥 275 pages
Why a Book About Claude Code?
Claude Code changed how I build software. It is not a suggestion engine — it is a tool-use engine that can read files, run commands, control a browser, review a pull request, and hook into your CI/CD pipeline. The gap I kept seeing in teams is not that they lack access to a good model; it is that they never move beyond autocomplete. They prompt, copy, and paste, and stop.
Claude Code Projects closes that gap. It teaches the workflows I use daily to ship production systems, structured around ten real projects so every chapter ends in something you can run and reuse.
The Core Loop
Every project in the book follows the same five-step loop, which is also the loop I run in production with AI-assisted work:
- Plan — give Claude the task and the acceptance criteria before it writes code.
- Implement — delegate the work, then read what it produced.
- Test — verify the result against the criteria, not your gut.
- Review — treat every diff like a pull request from a colleague.
- Commit — ship only when the loop passes.
This loop is the spine of the book. It turns “AI that writes code” into “AI that participates in the full development lifecycle.”
What You Will Build
Ten projects, each producing a working deliverable:
- Claude Code workspace — set up a professional, repeatable development environment.
- CLAUDE.md context — build and maintain project context so Claude stops asking for the same information.
- Skills — package reusable, domain-specific capabilities that work across sessions and teams.
- Hooks — enforce security and quality automatically, before and after Claude acts.
- MCP servers — extend Claude beyond files and shells into the tools and data your project needs.
- Subagents and worktrees — run parallel, isolated work safely without clobbering each other.
- Testing and verification — test AI-generated code rigorously so “it works” means something.
- Git discipline — use the version-control safety net as the real guardrail, not blind trust.
- Least-privilege access — give Claude only the permissions a task needs.
- Capstone — build a production-grade API and ship it through a real pull request, with a full production-readiness go/no-go review.
The last project ties everything together: a capstone API built with Claude Code and shipped the way your team actually ships — through a reviewable pull request.
Why It Works: Skills, Hooks, and MCP
Three mechanisms do the heavy lifting, and they map directly onto how production systems are built:
- Skills let you encode repeatable, high-quality procedures that Claude applies on demand — the same idea as a well-tested runbook or a battle-tested playbook.
- Hooks are automated guardrails that run before or after a tool call. A pre-tool-use hook can block an action (for example, reading a
.envfile); a post-tool-use hook can run a type check and feed errors back for automatic correction. - MCP (Model Context Protocol) servers give Claude new capabilities — a browser, a database, a deployment pipeline — so it stops being just a code editor and becomes a full automation tool.
For teams already running AI workloads on Kubernetes and building agentic pipelines, this model maps naturally onto the Agent Client Protocol vision of composable, interoperable AI tooling.
The Companion Repository
Every project ships with code you can run. A companion GitHub repository includes the demos, the starting points, and reusable SKILL.md files — so you can apply the workflows from the book at work from day one, not just read about them.
Who Should Read This
Claude Code Projects is for:
- Developers who want to move past prompting and actually ship with Claude Code.
- Tech leads evaluating Claude Code for their teams and wanting a repeatable, reviewable process.
- Engineers building AI-assisted workflows who care about testing, security, and production readiness — not just speed.
Prerequisites: basic programming knowledge, Git familiarity, and a Claude Code account (Pro or Team plan).
Book Details
- Title: Claude Code Projects: Build 10 Real-World Projects with Skills, Hooks, MCP, Testing, Git, and Production Workflows
- Author: Luca Berton
- Publisher: Packt Publishing
- Release: December 9, 2026
- Pages: 275
- ISBN-13: 978-1809310064
- ISBN-10: 1809310061
- Formats: Ebook and print
📘 Pre-order Claude Code Projects on Amazon →
Why I Wrote It
After building AI platforms on Kubernetes, authoring eight technical books, and speaking at KubeCon and Red Hat Summit, I have seen the same pattern in nearly every team: they adopt an AI tool but never move beyond basic autocomplete. The difference between a tool that saves you a little time and one that changes how you ship is workflow discipline — planning, testing, reviewing, and reviewing the diff like a colleague’s pull request.
That is exactly what this book teaches, and it is the same discipline I use to build premium websites and automate complex deployments. If you are serious about using Claude Code in production, this is the book.
Related Content
- Claude Code: Complete Guide to Hooks, MCP, and the SDK
- Karpathy’s CLAUDE.md: AI Coding Discipline
- Agentic DevOps with Claude Code: Skills and Subagents
- Claude Code Modifications: A Production-Ready Workflow
About the Author
I am Luca Berton, a Production AI Expert and author. I help enterprises integrate AI tools into their development workflows for measurable productivity gains. Claude Code Projects is my latest book with Packt Publishing. Book a consultation.
