Claude Code
Claude Code is Anthropic’s official CLI and VS Code extension for AI-powered software development. It gives Claude direct access to your codebase — reading files, writing code, running commands, and managing Git — all from your terminal or IDE.
What Makes Claude Code Different
Unlike chat-based AI assistants, Claude Code operates inside your development environment with full filesystem access. It doesn’t just suggest code — it reads your entire project, understands the architecture, makes targeted changes, runs tests, and commits to Git.
| Capability | Description |
|---|---|
| Codebase awareness | Reads and indexes your entire project structure |
| File operations | Creates, edits, and deletes files directly |
| Command execution | Runs shell commands (build, test, deploy) |
| Git integration | Commits, branches, creates PRs |
| MCP connections | Connects to databases, APIs, external tools |
| Multi-session | Parallel instances for different features |
Key Concepts at a Glance
CLAUDE.md
Your project’s brain — persistent instructions Claude reads at every session start. Controls behavior, conventions, and project context.
Skills
Modular instruction packages that load on demand. 90+ available skills across document handling, design, development, and enterprise workflows.
MCP Servers
Connect Claude to external systems — databases, APIs, GitHub, Slack, Notion — through the Model Context Protocol.
Worktrees
Run multiple Claude sessions in parallel on different feature branches using Git worktrees.
Models Available
| Model | ID | Best For |
|---|---|---|
| Opus 4.6 | claude-opus-4-6 | Complex tasks, architecture, deep reasoning |
| Sonnet 4.6 | claude-sonnet-4-6 | Routine development, balanced speed/quality |
| Haiku 4.5 | claude-haiku-4-5-20251001 | Quick tasks, trivial operations |
The PSB Workflow — Plan, Setup, Build
This structured methodology separates effective Claude Code users from beginners.
1. PLAN — Before Any Code
2. SETUP — Environment Checklist
.env with API keys, CLAUDE.md with project rules, docs/ folder structure, .claude/settings.json with permissions, install relevant MCPs and skills.3. BUILD — Feature by Feature
Permission Modes
| Mode | Behavior | Toggle |
|---|---|---|
| Default | Asks permission for every file change or command | Shift+Tab |
| Accept Edits | Edits files autonomously, may still ask for commands | Shift+Tab |
| Plan Mode | Read-only — thinks and plans, writes zero code | Shift+Tab |
| Bypass Permissions | Auto-accepts ALL tool calls (YOLO mode) | --dangerously-skip-permissions flag |
Essential Commands
| Command | Purpose |
|---|---|
/model | Switch between Opus, Sonnet, Haiku |
/init | Scan codebase and generate CLAUDE.md |
/memory | Edit persistent rules |
/compact | Compress conversation to save tokens |
/clear | Fresh start for unrelated tasks |
/context | Visualize token usage |
/mcp | Manage MCP server connections |
/fast | Toggle fast output mode |
The 10 Mantras
- Read first, answer second — Never guess when you can verify
- Simplicity is sophistication — The best code is the least code
- Make it work, make it right, make it fast — In that order
- Convention over configuration — Follow established patterns
- Fail fast, fail loud — Errors should be obvious, not silent
- Measure, don’t assume — Profile before optimizing
- Security is not optional — Build it in from the start
- Documentation is code — Keep them in sync
- User success is my success — Focus on outcomes
- Excellence is a habit — Consistent quality, not occasional brilliance