Claude Code Cheat Sheet

This page condenses Claude Code into two pages: install, headless mode, slash commands, shortcuts, CLAUDE.md, permissions, settings, hooks, skills, subagents, and MCP. You can also download the information as a printable cheat sheet:

Free Bonus: Claude Code Cheat Sheet

Get the Claude Code Cheat Sheet (PDF), a printable reference to install and launch flags, headless mode, slash commands, shortcuts, CLAUDE.md, permissions, hooks, skills, subagents, and MCP:

Claude Code Cheat Sheet

Want a guided path instead of a reference? Start learning at realjs.com.

Install and Start

  • Start claude in your project root and avoid your home folder.
  • Sign in with a paid Claude plan or an Anthropic Console account.
  • Type / for commands and ? for shortcuts.
Shell Install on macOS and Linux 
$ curl -fsSL https://claude.ai/install.sh \
    | bash
Windows PowerShell Install on Windows 
PS> irm https://claude.ai/install.ps1 | iex
Shell Check and Update 
$ claude --version
$ claude update
Shell Start a Session 
$ cd my-project/
$ claude                      # Interactive
$ claude "explain this repo"  # With prompt
$ claude -n auth-fix          # Named
Shell Resume a Session 
$ claude -c          # Latest one in this dir
$ claude -r          # Pick from a list
$ claude -r auth-fix "finish the PR"
Shell Launch Flags 
$ claude --model sonnet
$ claude --effort high
$ claude --permission-mode plan
$ claude --add-dir ../shared-lib
$ claude -w feature-x   # In a git worktree

Run Claude Code Headless

  • -p prints one answer and exits, so it fits scripts and CI.
  • --allowedTools pre-approves the tools a script needs.
  • --bare skips CLAUDE.md, hooks, skills, and MCP for faster starts.
Shell One-Shot Prompts 
$ claude -p "explain this function"
$ cat error.log | claude -p "root cause?"
$ claude -c -p "check for type errors"
Shell JSON Output and Limits 
$ claude -p "summarize this project" \
    --output-format json | jq -r '.result'
$ claude -p "fix the failing test" \
    --allowedTools "Read,Edit,Bash" \
    --max-turns 5

Session Slash Commands

  • Type / to list commands, then keep typing to filter.
  • Put the command at the start of your message.
  • Run /clear when you switch tasks and /compact to keep going.

Context

Command What it does
/context Show context usage
/compact [focus] Summarize the chat to free space
/clear Start a fresh chat. Also /new
/btw [question] Ask a side question

Sessions

Command What it does
/resume [name] Reopen a session. Also /continue
/rename [name] Label this session
/branch [name] Try another direction
/rewind Roll back code and chat. Also /undo
/export [file] Save the chat as plain text
/exit Quit. Also /quit

Background Work

Command What it does
/tasks List running shells and agents
/subtask <task> Hand a task to a forked subagent
/fork [prompt] Run a copy in the background
/background Detach the session. Also /bg
/goal [condition] Keep working until the goal is met
/loop 5m <prompt> Repeat a prompt on an interval

Setup and Review Slash Commands

  • Press s in the /model picker to switch for this session only.
  • Raise /effort for hard problems, at the cost of more tokens.

Model and Usage

Command What it does
/model [name] Switch model and save it as default
/effort high Set effort from low to max, or auto
/fast Toggle fast mode, or pass on or off
/usage Show cost and limits. Also /cost

Configure

Command What it does
/config Open settings. Also /settings
/config theme=dark Set one key directly
/permissions Edit allow, ask, and deny rules
/memory Edit memory files
/hooks View configured hooks
/skills List and toggle skills
/mcp Manage MCP servers and OAuth
/doctor Diagnose your setup
/status Show version and account

Plan and Review

Command What it does
/init Draft a CLAUDE.md
/plan [task] Enter plan mode
/diff Review the working tree
/code-review --fix Find and fix bugs
/code-review ultra Run a deep cloud review
/security-review Find security issues
/simplify Clean up the code

More Slash Commands

  • Your skills and plugin skills share the same / menu.
  • MCP server prompts show up as /server:prompt.

Workspace

Command What it does
/add-dir <path> Grant access to another directory
/cd <path> Move the session and keep the chat
/ide Connect your IDE
/copy [N] Copy a response
/insights Report on your sessions

Look and Feel

Command What it does
/theme Pick a color theme
/color [color] Set the prompt bar color
/statusline Configure the status line
/keybindings Edit the shortcuts file
/help List all commands
/release-notes See what’s new

Extend

Command What it does
/plugin Browse and install plugins
/reload-skills Pick up new skills
/workflows Watch dynamic workflows
/schedule Set up recurring cloud routines

Keyboard Shortcuts

  • On macOS, press Option for Alt. Most of these need Option set as Meta, but Option+T works without it.

Control the Session

Keys Action
Shift+Tab Cycle permission modes
Esc Interrupt Claude
Esc Esc Clear input or rewind
Ctrl+C Interrupt, clear, or exit when pressed twice
Ctrl+O Toggle transcript viewer
Ctrl+B Background a running task
Ctrl+T Toggle the task checklist
Alt+P Switch model
Alt+T Toggle extended thinking
Alt+O Toggle fast mode
Ctrl+L Redraw the screen

Write Your Prompt

Keys Action
\ Enter, Ctrl+J New line
Shift+Enter New line in most terminals
Ctrl+R Search prompt history
Ctrl+G Edit prompt in your editor
Ctrl+S Stash or restore the prompt
Ctrl+V Paste an image
Cmd+V Paste an image in iTerm2
Alt+V Paste an image on Windows or WSL
! at start Shell mode
@ Mention a file or subagent
? Show all shortcuts

CLAUDE.md and Memory

  • CLAUDE.md holds instructions you write. Auto memory holds notes Claude writes.
  • All files load together from broad to specific, so remove contradictions.
  • Since v2.1.277, AGENTS.md loads when a repo has no CLAUDE.md.

Where CLAUDE.md Files Live

File Scope
~/.claude/CLAUDE.md You, in every project
./CLAUDE.md Your team. Commit it
./CLAUDE.local.md You, in this project
sub/CLAUDE.md Loads on demand
Markdown Text Import Files 
# Project Notes
See @README.md for an overview.
- Testing: @docs/testing.md
- Mine: @~/.claude/my-notes.md
Markdown Text Scope a Rule in .claude/rules/api.md 
---
paths:
  - "src/api/**/*.js"
---
- Validate every request body

Permissions and Modes

  • Shift+Tab cycles default, acceptEdits, and plan, plus auto when it’s available.
  • Since v2.1.283, most terminal and VS Code sessions start in auto.
  • The UI labels default mode as Manual.
  • Rules check deny, then ask, then allow. The first match wins.
  • Deny rules hold in every mode, even bypassPermissions.
Mode Runs without asking
default Reads only
acceptEdits Reads, edits, file commands
plan Reads to draft a plan
auto Everything, with safety checks
dontAsk Pre-approved tools and file reads
bypassPermissions Everything, in isolated VMs
JSON Permission Rules 
{
  "permissions": {
    "allow": ["Bash(npm test *)"],
    "ask": ["Bash(git push *)"],
    "deny": ["Read(./.env)"]
  }
}

Settings Files

  • Managed settings win, followed by --settings, local, project, and user.
  • List keys like permissions.allow merge across files.
  • Settings files are strict JSON with no comments or trailing commas.
  • /status shows which files loaded.

Where Settings Live

File Scope
~/.claude/settings.json User
.claude/settings.json Project, shared
.claude/settings.local.json Project, just you

Hooks

  • Hooks run your commands at fixed points in Claude’s loop.
  • Exit code 2 from PreToolUse blocks the tool call.
  • /hooks shows what’s registered.

Key Events

Event When it runs
SessionStart Session begins or resumes
UserPromptSubmit Before Claude sees your prompt
PreToolUse Before a tool runs. Can block it
PostToolUse After a tool succeeds
Notification Claude needs your input
Stop Claude finished replying
JSON Format After Every Edit 
{
  "hooks": {
    "PostToolUse": [{
      "matcher": "Edit|Write",
      "hooks": [{
        "type": "command",
        "command": "npx prettier --write ."
      }]
    }]
  }
}

Skills

  • A folder with a SKILL.md becomes a /name command.
  • Claude also loads a skill when its description fits the task.
  • disable-model-invocation: true makes a skill manual-only.
  • Old .claude/commands/*.md files still work.

Where Skills Live

File Scope
~/.claude/skills/<name>/SKILL.md You
.claude/skills/<name>/SKILL.md Project
Markdown Text A Minimal SKILL.md 
---
name: jsdoc-writer
description: >-
  Writes JSDoc comments for functions.
  Use when asked to document code.
---
Open with a one-line summary...

Claude follows the instructions below the frontmatter each time the skill runs.

Subagents

  • A subagent works in its own context and reports back.
  • Built-in subagents include Explore, which searches without editing, and Plan.
  • Claude can write one for you, or you can add a file to .claude/agents/.
Markdown Text Define .claude/agents/code-reviewer.md 
---
name: code-reviewer
description: Reviews code for quality
tools: Read, Glob, Grep
model: sonnet
---
You are a code reviewer...
Shell Run the Session as an Agent 
$ claude --agent code-reviewer
Text Call a Subagent 
@agent-code-reviewer check my changes

MCP Servers

  • MCP connects Claude Code to outside tools and data.
  • Scopes are local, project, and user, with local as the default.
  • Project scope saves servers to a shared .mcp.json file.
  • For stdio servers, -- comes before the server command.
  • Connect databases with a read-only user.
Shell Add a Server 
$ claude mcp add --transport http \
    notion https://mcp.notion.com/mcp
$ claude mcp add --transport stdio \
    --scope project db \
    -- npx -y @bytebase/dbhub \
    --dsn "postgresql://readonly:pw@host/db"
Shell Manage Servers 
$ claude mcp list
$ claude mcp get notion
$ claude mcp remove notion

Everyday Workflow

  • Switch to plan mode with Shift+Tab before big changes.
  • Review every diff and commit between tasks.
  • Run /branch before a risky experiment.
  • /rewind restores Claude’s file edits but not changes from shell commands.

A Typical Loop

Command What it does
/init Once per project
/plan add a CLI Agree on a plan
/diff Review the edits
/code-review --fix Catch bugs early
/clear Start the next task

You can download this information as a printable cheat sheet:

Free Bonus: Claude Code Cheat Sheet

Get the Claude Code Cheat Sheet (PDF), a printable reference to install and launch flags, headless mode, slash commands, shortcuts, CLAUDE.md, permissions, hooks, skills, subagents, and MCP:

Claude Code Cheat Sheet