Claude Code Cheatsheet

Quick reference for Claude Code CLI commands, shortcuts, and configuration

Claude Code is Anthropic’s official CLI tool for interacting with Claude directly from your terminal. This cheatsheet covers the essential commands, shortcuts, and configuration options.

Installation & Setup

# Install via npm
npm install -g @anthropic-ai/claude-code

# Launch Claude Code
claude

# First run will prompt for Anthropic API key or OAuth login

Essential Keyboard Shortcuts

Shortcut	Action
Enter	Send message
Shift+Enter	New line in input
Ctrl+C	Cancel current operation
Ctrl+L	Clear screen
Esc	Interrupt Claude / Exit input mode
Up/Down	Navigate command history
Tab	Autocomplete file paths

Slash Commands

Type these directly in the chat:

Command	Description
/help	Show help and available commands
/clear	Clear conversation history
/compact	Condense conversation to save context
/config	Open configuration settings
/cost	Show token usage and cost for session
/doctor	Diagnose installation issues
/init	Initialize Claude Code in current project
/login	Switch accounts or re-authenticate
/logout	Log out of current account
/memory	View/edit persistent memory (CLAUDE.md)
/model	Switch between available models
/permissions	Manage tool permissions
/review	Request code review
/status	Show current session status
/tasks	View running background tasks
/vim	Toggle vim keybindings

CLI Flags

# Start with a specific prompt
claude "explain this codebase"

# Continue last conversation
claude --continue
claude -c

# Resume specific conversation
claude --resume <conversation_id>

# Print output only (no interactive mode)
claude --print "what does main.py do?"
claude -p "what does main.py do?"

# Use specific model
claude --model claude-sonnet-4-20250514

# Pipe input to Claude
cat error.log | claude "explain this error"

# Run in headless/non-interactive mode
claude --headless "fix the bug in auth.py"

# Output JSON format
claude --output-format json -p "list files"

# Set max turns for agentic loops
claude --max-turns 10 "refactor this module"

# Allow specific tools without prompting
claude --allowedTools "Bash(npm:*),Read,Write"

# Disable specific tools
claude --disallowedTools "Bash(rm:*)"

Configuration Files

CLAUDE.md (Project Memory)

Create CLAUDE.md in your project root to give Claude persistent context:

# Project: MyApp

## Tech Stack
- Python 3.11, FastAPI, PostgreSQL
- Frontend: React + TypeScript

## Conventions
- Use snake_case for Python, camelCase for JS
- All API endpoints go in `src/api/`
- Tests mirror source structure in `tests/`

## Common Commands
- `make dev` - Start development server
- `make test` - Run test suite

Settings Location

# Global settings
~/.claude/settings.json

# Project-level settings
.claude/settings.json

# Project-level settings (local, not committed)
.claude/settings.local.json

Permission Modes

Control how Claude asks for permission:

Mode	Behavior
default	Ask before file writes and shell commands
plan	Research only, no code changes
trust-tools	Allow all tools without asking

# Set via flag
claude --permission-mode trust-tools

# Or in settings.json
{
  "permissions": {
    "defaultMode": "default"
  }
}

MCP Servers

Model Context Protocol (MCP) servers extend Claude’s capabilities:

// ~/.claude/settings.json
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/dir"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "your-token"
      }
    }
  }
}

Popular MCP servers: - @modelcontextprotocol/server-filesystem - Enhanced file operations - @modelcontextprotocol/server-github - GitHub API integration - @modelcontextprotocol/server-postgres - Database queries - @modelcontextprotocol/server-brave-search - Web search

Hooks

Run custom scripts on Claude events:

// .claude/settings.json
{
  "hooks": {
    "onFileWrite": [
      {
        "command": "prettier --write",
        "match": "*.{js,ts,json}"
      }
    ],
    "onSessionStart": [
      {
        "command": "echo 'Session started'"
      }
    ]
  }
}

Available hooks: - onFileWrite - After file is written - onFileRead - After file is read - onSessionStart - When session begins - onSessionEnd - When session ends - onPromptSubmit - Before prompt is sent

Skills & PluginsSkills and plugins extend Claude Code with reusable prompts and third-party integrations.### User SkillsCreate custom slash commands by adding markdown files to ~/.claude/skills/:markdown<!-- ~/.claude/skills/commit.md -->---description: "Generate a git commit with conventional format"---Review the staged changes and create a commit:1. Analyze `git diff --staged`2. Write a conventional commit message (feat/fix/docs/refactor/test)3. Keep the subject line under 50 characters4. Add body explaining "why" not "what"Skill file structure:~/.claude/skills/├── commit.md → /commit├── review-pr.md → /review-pr├── explain.md → /explain└── subfolder/ └── deploy.md → /subfolder:deploySkill frontmatter options:yaml---description: "Short description shown in /help" # Requiredargs: "optional arguments description" # Optional---### Project-Level SkillsAdd skills specific to a project in .claude/skills/:markdown<!-- .claude/skills/test.md -->---description: "Run project tests with coverage"---Run the test suite for this project:1. Execute `npm test -- --coverage`2. If tests fail, analyze failures and suggest fixes3. Report coverage summary### Installing PluginsPlugins are community-contributed skill packs. Install via the registry:bash# Install a pluginclaude /install-plugin @username/plugin-name# Or add manually to settings.json``````json// ~/.claude/settings.json{ "plugins": [ "ralph-wiggum@claude-plugins-official", "frontend-design@claude-plugins-official" ]}### Popular Plugins| Plugin | Description ||——–|————-|| frontend-design | Create polished UI components || ralph-wiggum | Agentic loop for complex multi-step tasks |### Plugin Skills UsagePlugin skills are invoked with the plugin prefix:bash# Use frontend-design plugin/frontend-design:frontend-design# Use ralph-wiggum plugin /ralph-wiggum:ralph-loop/ralph-wiggum:cancel-ralph/ralph-wiggum:help### Skill Best Practices1. Be specific - Include exact commands and expected outputs2. Add context - Tell Claude what files/patterns to look for3. Chain actions - Skills can include multiple steps4. Use variables - Reference $ARGS for user inputmarkdown<!-- ~/.claude/skills/scaffold.md -->---description: "Scaffold a new component"args: "<component-name>"---Create a new React component named $ARGS:1. Create `src/components/$ARGS/$ARGS.tsx`2. Create `src/components/$ARGS/$ARGS.test.tsx`3. Create `src/components/$ARGS/index.ts` barrel export4. Use TypeScript and follow project conventions from CLAUDE.md

Pro Tips

1. Use /compact liberally - Keeps context window efficient for long sessions

2. Create project-specific CLAUDE.md - Claude reads this automatically and follows your conventions

3. Pipe content directly - git diff | claude "review these changes" is powerful

4. Use --print for scripts - Get clean output for automation

5. Check /cost periodically - Monitor your API usage

6. Run /doctor - When things aren’t working right

7. Use background tasks - Claude can run long operations without blocking

8. Trust but verify - Always review code changes before committing

Useful Patterns

# Quick code explanation
claude -p "explain src/auth/" | less

# Debug with error logs
tail -100 app.log | claude "what's causing these errors?"

# Code review workflow
git diff main | claude "review this PR"

# Generate commit message
git diff --staged | claude -p "write a commit message"

# Explore unfamiliar codebase
claude "give me an overview of this project's architecture"

# Fix failing tests
npm test 2>&1 | claude "fix these test failures"

Resources

• Claude Code Documentation
• GitHub Repository
• MCP Specification