Why Claude Code?
Delegating real engineering work to an agent, not just autocompleting lines
Most coding assistants finish your sentence. Claude Code is built to finish your task — you describe what you want in plain English, and it reads the codebase, makes a plan, writes and edits files, runs tests, and iterates until the change actually works. That shift, often called "vibe coding," changes what's worth learning: less about memorizing syntax, more about giving a capable agent the right context, the right guardrails, and the right amount of autonomy for the job at hand.
It's also a tool with real depth once you get past the basics — CLAUDE.md files that persist project context, skills and subagents that scope what the agent knows and does, hooks that automate your workflow around it, and MCP servers that connect it to the rest of your stack. Whether you're using it through the CLI, an IDE extension, or the desktop app, the roadmap below covers the concepts that turn "asking Claude to code something" into a repeatable, scalable workflow.
Quick intro — what is Claude Code?
A quick primer before you start the roadmap. Opens in a small player, no need to leave the page.
The Claude Code Roadmap
Work through these in order. Each step has a short lesson, official docs, and the commands you'll actually use.
Vibe Coding & Coding Agents
What vibe coding means, what makes Claude Code a coding agent rather than an autocomplete tool, and the different ways to access Claude (subscription vs API).
Claude CLI & Session Commands
Install the CLI, run claude and claude "query", and get comfortable with sessions: -p, -c, -r, and the shortcuts that keep you moving.
CLAUDE.md, Skills & Tools
Give Claude persistent project context with CLAUDE.md, package up reusable instructions as Skills, and understand the built-in tools Claude can call.
Models & Context Management
Know when to reach for Opus vs Sonnet vs Haiku, and how to keep a long session healthy with /compact, /clear, and /context.
Slash Commands & Common Usecases
The command cheatsheet — /help, /status, /usage, /cost, /init, /doctor — plus the workflows people actually reach for Claude Code to do.
Permission Modes, Plan Mode & Sessions
Control what Claude can do without asking, use Plan Mode to review before it acts, and resume or fork sessions instead of starting over.
Subagents & Agent Teams
Delegate focused work to subagents with their own context and tools, and coordinate multiple agents at once with agent teams.
Hooks & Automation
Wire in hooks like PreToolUse, PostToolUse, and SessionStart to auto-format code, block risky edits, and automate everything around the agentic loop.
MCP & Plugins
Connect Claude Code to external tools and data sources with MCP servers, and package skills, hooks, and MCP configs together as installable plugins.
Scaling Claude
Headless mode and non-interactive automation, git worktrees for parallel work, pricing and prompt caching, and security best practices as usage grows.
Practice Exercises
Real ways to build muscle memory with the tool, not just read about it
Write a Real CLAUDE.md
Run /init in one of your own repos, then hand-edit the generated CLAUDE.md until it reflects how your team actually works.
Build a Custom Skill
Package a repeatable task you do often — like writing a changelog entry or scaffolding a component — as a reusable Claude Code skill.
Set Up an Auto-Format Hook
Add a PostToolUse hook that runs your linter or formatter automatically every time Claude edits a file.
Connect an MCP Server
Connect a real MCP server — GitHub, a database, or a monitoring tool — and have Claude Code use it mid-session.
Track complete
Ten steps, from your first claude command to scaling with subagents, hooks, and MCP. Docs move fast here — keep the official Claude Code documentation bookmarked and revisit it as features ship.
Where next?
Keep exploring by domain or drill into a single skill