Study Plan — Claude Code
For developers who want to go from zero to deep understanding of Claude Code, with universal concepts transferable to any other code agent (Cursor, Copilot, Aider, etc.).
Version: April/2026 · Primary focus: Claude Code (Anthropic) · Content license: CC BY 4.0
Why Claude Code as the focus
Claude Code is IDE-agnostic (terminal-first), has first-class MCP, subagents, hooks, slash commands, skills, plugins and its own SDK — covering the full spectrum from beginner to engineer building agents. The concepts you learn here (tool use, context engineering, MCP, evals, multi-agent) are transferable to any other agent on the market.
Other tools (Cursor, Copilot, Aider, OpenHands) appear as occasional comparative references — never as the primary topic.
How to use this plan
The plan is linear by default (Phase 1 → 2 → 3 → 4 → Capstone), but designed so you can progress through the phase that makes most sense for your moment. Each module indicates which profiles it is essential for.
Tracks by profile
| Track | Profile | Focus | Suggested path |
|---|---|---|---|
| 🟢 Productivity | Junior/mid dev | Use agents well day-to-day | Phase 1 → 2 → selected modules from 3 → light Capstone |
| 🔵 Tech leadership | Senior dev / tech lead | 🟢 track + adoption, metrics, governance | Everything, with extra weight on Phase 4 (M23–M26) |
| 🟣 Platform / AI | Platform or AI engineer | 🔵 track + building agents, MCPs and SDK | Everything, with extra weight on Phases 3 and 4 (M13–M16, M19–M22, M27) |
In each module below, 🟢🔵🟣 markings indicate which tracks it is essential for. When a track does not appear, the module is optional for it.
Material marking conventions
- Type: 📚 Official doc · 📖 Article/Paper · 🎓 Course · 🎬 Video/Podcast · 🛠️ Hands-on
- Language:
[PT]·[EN] - Cost:
[Free]·[Paid]·[Freemium] - Level:
beginner·intermediate·advanced - ⭐ essential · 🌐 universal concept (transferable to any agent)
Overall map
| Phase | Theme | Duration | Dedication |
|---|---|---|---|
| 1 | Fundamentals — understanding agents | 3 weeks | 6–8h/week |
| 2 | Fluency — daily use | 4 weeks | 8–10h/week |
| 3 | Advanced — extending the agent | 4 weeks | 10–12h/week |
| 4 | Mastery — building, scaling, governing | 5 weeks | 8–10h/week |
| Capstone | Integrative project | 1–2 weeks | 10–15h/week |
Full path total: 18–20 weeks (~4 to 5 months). 🟢 lean track: ~10–12 weeks.
Complete structure (27 modules)
Phase 1 — Fundamentals
M1 Market overview and why Claude Code · M2 Installation and setup · M3 Prompt engineering and context engineering · M4 First hands-on workflows
Phase 2 — Fluency
M5 Automation · M6 Spec-driven development · M7 TDD with Claude Code · M8 Git/PR/code review · M9 Debugging · M10 Automated documentation · M11 Large codebases
Phase 3 — Advanced
M12 Headless and CI/CD · M13 MCP · M14 Subagents · M15 Hooks and slash commands · M16 Skills and plugins · M17 Context and memory · M18 Security and prompt injection
Phase 4 — Mastery
M19 Claude Agent SDK · M20 Multi-agent · M21 Evals · M22 Costs and tokens · M23 Metrics and ROI · M24 Legal and IP · M25 Enterprise governance · M26 Culture and adoption · M27 When NOT to use Claude Code
Capstone — 3 variants by track
- 🟢 Light capstone: complete personal plugin + CI + sandbox + optimized costs
- 🔵 Team capstone: 4-week pilot with governance, metrics and retrospective
- 🟣 Product capstone: custom agent with Agent SDK + own MCP + evals + publication
How to contribute
This plan is open content (CC BY 4.0). Pull requests with corrections, link updates and new materials are welcome. Open an issue or send a PR with suggestions.
Ready to start? Go to Phase 1 →