Claude Code Cheat Sheet

Name: Claude Code
Author: Anthropic

Installation & Quick Start

Installation

npm Install	`npm install -g @anthropic-ai/claude-code`
Run	`claude`
Update	`npm update -g @anthropic-ai/claude-code`
Version	`claude --version`

Requires Node.js 18+. First run requires Anthropic account authentication.

Authentication

Login	`/login`
Logout	`/logout`
Use API Key	`ANTHROPIC_API_KEY=sk-xxx claude`
Account Status	`/status`

Get API key from Anthropic Console, or use Claude Pro/Max subscription.

Basic Usage

Interactive Mode

Start	`claude`
Specific Directory	`claude /path/to/project`
Continue Last	`claude -c`
Select History	`claude -r`

One-liner Mode

Single Command	`claude -p "Write tests"`
JSON Output	`claude -p "..." --output-format json`
Pipe Input	`cat file.ts \| claude -p "Review this"`
Streaming	`claude -p "..." --output-format stream-json`

Output Format

`text`	Default text output
`json`	JSON format (after completion)
`stream-json`	Real-time JSON stream

Permission Mode

`--allowedTools`	Allow specific tools only
`--disallowedTools`	Block specific tools
`--dangerouslySkipPermissions`	Skip all permission checks

Use --dangerouslySkipPermissions only in trusted environments!

Slash Commands

Session Management

`/clear`	Clear conversation
`/compact`	Summarize and compress context
`/status`	Check account & session status
`/cost`	Check current session cost

Work Commands

`/init`	Create CLAUDE.md project file
`/add-dir <path>`	Add working directory
`/model`	Change model
`/permissions`	Manage permission settings

File & Context

`/context`	View current context files
`/memory`	Edit CLAUDE.md memory file
`/review`	Request code review
`/pr-comments`	Handle PR comments

System

`/help`	Show help
`/config`	Open settings
`/bug`	Submit bug report
`/doctor`	Diagnose installation
`/quit`	Exit Claude Code

MCP Commands

`/mcp`	Check MCP server status
`/install-github-app`	Install GitHub app
`/ide`	IDE integration settings
`/terminal-setup`	Terminal setup helper

Developer Tools

`/vim`	Toggle Vim mode
`/diff`	Change diff view mode
`/theme`	Change theme

Keyboard Shortcuts

Input & Navigation

macOS 🐧 Linux ⊞ Windows

Shortcut	Description	Shortcut	Description
Enter	Send message	Ctrl+L	Clear screen
Shift+Enter	New line	Ctrl+R	Search history
Ctrl+C	Cancel current operation	Ctrl+U	Delete entire line
Ctrl+D	End session	Ctrl+K	Delete to end of line
↑ / ↓	Previous/Next message	Tab	Autocomplete
Esc	Cancel current input

Shortcut	Description	Shortcut	Description
Return	Send message	⌘+K	Clear screen
⇧+Return	New line	⌃+R	Search history
⌃+C	Cancel current operation	⌃+U	Delete entire line
⌃+D	End session	⌃+K	Delete to end of line
↑ / ↓	Previous/Next message	Tab	Autocomplete
Esc	Cancel current input

Shortcut	Description	Shortcut	Description
Enter	Send message	Ctrl+L	Clear screen
Shift+Enter	New line	Ctrl+R	Search history
Ctrl+C	Cancel current operation	Ctrl+U	Delete entire line
Ctrl+D	End session	Ctrl+K	Delete to end of line
↑ / ↓	Previous/Next message	Tab	Autocomplete
Esc	Cancel current input

Permission Response

y	Allow this time
n	Deny
a	Always allow (session)
!	Allow once, no retry

Vim Mode (when enabled)

i	Insert mode
Esc	Normal mode
v	Visual mode
:	Command mode

Toggle Vim mode with /vim command.

CLI Options

Main CLI Flags

Option	Short	Description
`--print`	`-p`	Run single prompt in non-interactive mode
`--continue`	`-c`	Continue most recent conversation
`--resume`	`-r`	Resume previous conversation with selector
`--model`	`-m`	Specify model (e.g., claude-sonnet-4-20250514)
`--output-format`		Output format: text, json, stream-json
`--max-turns`		Max turns limit in non-interactive mode
`--system-prompt`		Custom system prompt
`--allowedTools`		Allowed tools (comma-separated)
`--disallowedTools`		Disallowed tools (comma-separated)
`--permission-mode`		Permission mode: default, plan, bypassPermissions
`--dangerously-skip-permissions`		⚠️ Skip all permission checks
`--verbose`		Verbose logging

Configuration

Config File Paths

macOS 🐧 Linux ⊞ Windows

Global Settings	`%USERPROFILE%\.claude\settings.json`
Project Settings	`.claude\settings.json`
Local Settings	`.claude\settings.local.json`

Global Settings	`~/.claude/settings.json`
Project Settings	`.claude/settings.json`
Local Settings	`.claude/settings.local.json`

Global Settings	`~/.claude/settings.json`
Project Settings	`.claude/settings.json`
Local Settings	`.claude/settings.local.json`

Priority: Local > Project > Global. Add local settings to .gitignore.

Config Example

{
  "model": "claude-sonnet-4-20250514",
  "permissions": {
    "allow": ["Bash(*)", "Read(*)", "Write(*)"],
    "deny": []
  }
}

Environment Variables

`ANTHROPIC_API_KEY`	Set API key
`CLAUDE_MODEL`	Set default model
`CLAUDE_CONFIG_DIR`	Config directory path
`HTTP_PROXY`	Proxy server

Permission Patterns

`Bash(*)`	Allow all Bash commands
`Bash(npm:*)`	Allow npm commands only
`Read(*)`	Allow all file reads
`Write(src/*)`	Allow write to src/ only
`mcp__server__tool`	Allow specific MCP tool

Skills

What are Skills?

Definition	Custom commands that extend Claude via SKILL.md files
Invocation	Invoke with /skill-name or Claude applies automatically
Components	YAML frontmatter + markdown instructions
Supporting files	Can include templates, examples, scripts

Skills extend Claude capabilities. Create SKILL.md with instructions, Claude adds it to its toolkit.

Skill Locations

macOS 🐧 Linux ⊞ Windows

Global Skills	`%USERPROFILE%\.claude\skills\<name>\SKILL.md`
Project Skills	`.claude\skills\<name>\SKILL.md`

Global Skills	`~/.claude/skills/<name>/SKILL.md`
Project Skills	`.claude/skills/<name>/SKILL.md`

Global Skills	`~/.claude/skills/<name>/SKILL.md`
Project Skills	`.claude/skills/<name>/SKILL.md`

Each skill is a directory with SKILL.md as the entrypoint. Supports templates, examples, and scripts.

SKILL.md Frontmatter

`name`	Skill name (becomes /slash-command)
`description`	Description for when Claude should apply
`disable-model-invocation`	If true, only user can invoke (no auto)
`user-invocable`	If false, hidden from / menu
`allowed-tools`	Tools allowed when skill is active
`context`	fork: run in subagent
`agent`	Agent type (Explore, Plan, etc.)

Skill Example

# Skill directory structure
~/.claude/skills/my-skill/
├── SKILL.md          # Required: main instructions
├── template.md       # Optional: template
└── scripts/
    └── helper.py     # Optional: script

# SKILL.md example
---
name: code-review
description: Review code and provide feedback
disable-model-invocation: false
allowed-tools: Read, Grep
---

Review the code changes:
1. Check code quality
2. Find bugs/issues
3. Suggest improvements

Skill Invocation

Direct invoke	`/skill-name` or `/skill-name args`
Auto apply	Claude auto-loads when description matches
`$ARGUMENTS`	Substitutes arguments passed at invocation
!`command`	Run shell command, insert output

Skills vs Commands

Slash commands	Single .md file, /... autocomplete in terminal
Skills	Directory structure, supporting files, Claude auto-apply
Compatibility	.claude/commands/ files still work
Priority	Skills take precedence over same-name commands

MCP Servers

MCP Config File

macOS 🐧 Linux ⊞ Windows

Global	`%USERPROFILE%\.claude\mcp_servers.json`
Project	`.claude\mcp_servers.json`

Global	`~/.claude/mcp_servers.json`
Project	`.claude/mcp_servers.json`

Global	`~/.claude/mcp_servers.json`
Project	`.claude/mcp_servers.json`

MCP (Model Context Protocol) servers extend Claude Code functionality.

MCP Config Example

{
  "mcpServers": {
    "github": {
      "command": "npx",
      "args": ["-y", "@anthropic-ai/mcp-server-github"],
      "env": { "GITHUB_TOKEN": "${GITHUB_TOKEN}" }
    }
  }
}

Popular MCP Servers

`server-github`	GitHub API integration
`server-filesystem`	Extended filesystem access
`server-postgres`	PostgreSQL database
`server-puppeteer`	Browser automation
`server-slack`	Slack integration

MCP Commands

`/mcp`	List connected MCP servers
`claude mcp add`	Add MCP server
`claude mcp remove`	Remove MCP server
`claude mcp list`	List configured servers

Hooks

Hook Events

`PreToolUse`	Before tool execution
`PostToolUse`	After tool execution
`Notification`	On notification
`Stop`	On Claude response completion

Hook Config Example

{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Write",
      "hooks": [{
        "type": "command",
        "command": "npm run lint --fix $CLAUDE_FILE_PATH"
      }]
    }]
  }
}

Hook Environment Variables

`$CLAUDE_TOOL_NAME`	Current tool name
`$CLAUDE_TOOL_INPUT`	Tool input (JSON)
`$CLAUDE_FILE_PATH`	Target file path
`$CLAUDE_FILE_CONTENT`	File content

Hook Response

Exit 0	Success, continue
Exit 2	Block tool execution
stdout JSON	Send feedback to Claude

Output {"decision": "block", "reason": "reason"} to block tool execution.

Memory (CLAUDE.md)

CLAUDE.md Paths

macOS 🐧 Linux ⊞ Windows

Project Root	`.\CLAUDE.md`
Global	`%USERPROFILE%\.claude\CLAUDE.md`
Local (gitignore)	`.\CLAUDE.local.md`

Project Root	`./CLAUDE.md`
Global	`~/.claude/CLAUDE.md`
Local (gitignore)	`./CLAUDE.local.md`

Project Root	`./CLAUDE.md`
Global	`~/.claude/CLAUDE.md`
Local (gitignore)	`./CLAUDE.local.md`

Use /init command to create project CLAUDE.md.

CLAUDE.md Example

# Project Overview
This is a React + TypeScript web application.

## Build Commands
- Development: `npm run dev`
- Test: `npm run test`

## Notes
- Do not modify .env file

Memory Commands

`/memory`	Edit memory file
`/init`	Create project CLAUDE.md
"Remember this"	Claude saves to CLAUDE.md

Memory Priority

Priority 1	Current directory CLAUDE.md
Priority 2	Parent directory CLAUDE.md
Priority 3	Home directory CLAUDE.md
Merge	All file contents are merged

Tips & Tricks

Effective Prompts

Be specific	"Add error handling to login function"
Specify file	"Modify src/auth.ts file"
Step by step	"1. First analyze 2. Then fix"
Provide context	"This is an Express server"

Common Workflows

# Bug Fixing
"Fix this error: [error message]"

# Code Review
"Review this PR and suggest improvements"

# Write Tests
"Write unit tests for this function"

Productivity Tips

Attach images	Drag and drop screenshots
Share URL	Enter webpage URL directly
Compress context	`/compact` to save tokens
Continue chat	`claude -c` for quick resume

Troubleshooting

Auth error	`/logout` then login again
Slow response	`/compact` to clean context
MCP error	`/mcp` to check server status
Config issue	`/doctor` to diagnose

CI/CD Integration

# GitHub Actions Example
- name: Claude Code Review
  run: |
    npm i -g @anthropic-ai/claude-code
    claude -p "Review this PR" --output-format json
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_KEY }}

Security Best Practices

API key	Use environment variables only
Sensitive files	Exclude .env, credentials
Review permissions	Verify commands before allowing
Local settings	Use settings.local.json