Production-grade Claude Code setup — documented from real project usage.
Not prompts. Not tips. A complete agentic development system: agents, orchestration, memory, gates, rules, and every workflow your team needs — written down so it works the same way every time.
Four standalone playbooks. Each one is completely self-contained — pick the folder that matches your stack and never look at the others.
| Playbook | For |
|---|---|
| flutter-generic/ | Any Flutter project — Riverpod, Bloc, Provider, or anything else |
| flutter-getx/ | Flutter + GetX + Clean Architecture |
| laravel-generic/ | Any Laravel project |
| laravel-mvc/ | Laravel with our validated MVC setup |
Each playbook has its own README, template, and chapters. Open one folder. Everything you need is inside.
Templates — ready to copy. Every playbook ships with a complete template folder. Drop it into your project root and you have a working Claude Code agent setup immediately. No manual configuration from scratch.
Chapters — publishing progressively. Each playbook is backed by a detailed written guide covering why every piece of the setup exists, how to configure it correctly, and what breaks if you skip a step. Open any playbook folder to see what's currently available. Watch or star the repo to follow along as new chapters publish.
Every playbook documents the same core system — fully adapted for its framework:
- CLAUDE.md — the instruction file Claude reads at the start of every session
- settings.json — model selection, permissions, and environment configuration
- Hooks — automated triggers that fire before and after agent actions
- Rules system — layered rules that govern how every agent behaves
- Specialized agents — purpose-built agents for every stage of the development lifecycle
- Orchestra management — how agents are triggered, routed, and coordinated
- MCP integration — cross-session memory so Claude doesn't start from scratch each time
- Token efficiency — how to avoid burning unnecessary tokens on repeated context loading
- Team collaboration — what to commit, what to gitignore, how shared memory works
- Mistakes and lessons — what went wrong in real usage and how it was fixed
1. Choose your playbook from the table above.
2. Open the playbook folder and read its README — it tells you exactly what's covered, who it's for, and what's currently published.
3. Copy the template into your project. Instructions are inside each playbook's README.
4. Follow the chapters as they publish to understand every decision behind the setup.
- You use Claude Code on a real project — not a tutorial app
- You've hit the wall: inconsistent output, no memory between sessions, agents that skip steps and improvise
- You want a system that produces the same quality every session — not a collection of clever prompts
This is not plug-and-play. It takes one focused session to configure. After that, it runs itself.
See CONTRIBUTING.md for how to report issues or suggest improvements.
Questions and discussion go in GitHub Discussions — not Issues.
MIT — copy, adapt, and use freely.