The complete cheat sheet

Show all available commands and keybindings.

Clear conversation history and start fresh. Context is wiped.

Compress conversation history to save context window space.

Open the Claude Code configuration editor. Manage models, permissions, and settings.

Show token usage and estimated cost for the current session.

Check Claude Code installation health. Verifies dependencies, auth, and configuration.

Exit Claude Code. Ctrl+D also works.

Generate a CLAUDE.md file. Claude analyses the codebase and creates project-specific instructions.

Authenticate with your Anthropic account via browser OAuth or API key.

Sign out from your Anthropic account and clear saved credentials.

Open memory files editor. Manage persistent notes that Claude remembers across sessions.

Select the AI model: Claude Opus 4.7, Sonnet 4.6, or Haiku 4.5.

View and manage tool permissions — files, bash, browser, etc.

Fetch and display comments from the current GitHub pull request.

View the Claude Code release notes and changelog for the installed version.

Request a code review. Claude analyses the diff and suggests improvements.

Show current account status, plan, and usage limits for the session.

Install shell integration (bash/zsh/fish). Enables Shift+Tab and other terminal features.

Toggle vim keybindings in the input prompt.

Open GitHub issue to report a Claude Code bug. Automatically includes version info.

Create a git commit with an AI-generated message based on staged changes.

Toggle fast output mode for quicker responses with Opus 4.7.

List connected MCP servers and their available tools and status.

Connect and manage IDE integrations (VS Code, JetBrains, Cursor). Enables real-time file sync.

Open the rewind UI — restore a past checkpoint, rolling back both chat history AND file changes. Also triggered by Esc+Esc.

Enter plan mode — read-only. Claude researches and proposes a plan without making file changes.

Set reasoning effort: low, medium, high, max, auto. Higher = more thinking tokens (Opus 4.7 only).

Open the interactive diff viewer showing all file changes made in the current session.

Visualize context window usage as a colored grid. Shows what's consuming space.

List and manage background tasks (subagents). View progress, outputs, stop tasks.

Run a prompt on a recurring interval. Use to poll status, monitor builds, or run periodic checks.

Ask a side question without adding it to conversation history.

Toggle voice dictation. Hold Space to record, release to transcribe. 20+ languages.

Open the keybindings editor (~/.claude/keybindings.json). Remap any shortcut or create chord sequences.

Add a directory to Claude's working context mid-session.

Show your current plan limits, rate limits, and usage quotas. More detailed than /cost.

Copy the last assistant response to clipboard. Pass N to copy the Nth-last response.

Rename the current session. Easier to resume with claude -r "name".

Export the current conversation as a text file.

View all configured hooks. Hooks run shell commands on events like PreToolUse, PostToolUse, SessionEnd.

Cancel the current operation or interrupt Claude mid-response.

Exit Claude Code (EOF signal). Same as /exit.

Auto-accept all pending edits. Cycles permission modes: default → acceptEdits → plan.

Accept current autocomplete suggestion in the input prompt.

Return to normal mode (vim) or dismiss autocomplete.

Enter rewind mode. Pick a past checkpoint to restore — rolls back conversation AND file changes.

Navigate through prompt history.

Clear the terminal screen.

Move cursor to beginning of the current input line.

Move cursor to end of the current input line.

Cut text from cursor to end of line. Ctrl+Y to paste back.

Cut text from cursor to start of line. Quick way to clear input.

Delete the word before the cursor.

Paste the last cut text back. Works with Ctrl+K and Ctrl+U cuts.

Toggle the background task list panel.

Toggle verbose transcript — shows all tool calls and their full inputs/outputs.

Open the current prompt in your $EDITOR for multi-line editing.

Background the current running task. Continue chatting while the previous task runs.

Toggle fast mode. Alt+O on Windows/Linux. Same as /fast.

Toggle extended thinking mode. ⌘+T on macOS, Meta+T on Linux/Windows.

Open the inline model picker. Quickly switch between Opus, Sonnet, and Haiku.

Jump word by word. Alt+←/→ on Windows/Linux.

Insert a newline without submitting. Write multi-line prompts.

In voice mode: hold Space to record, release to transcribe and send.

Stash the current prompt draft without sending. Retrieve it later in the same session.

Incremental reverse history search. Requires shell integration.

Run a shell command directly. Output appears in conversation so Claude can read it.

Add a note directly to Claude's memory. Persists across sessions.

Reference a file in your message. Claude reads and includes file content in context.

Forces structured reasoning. Dramatically improves accuracy on complex problems.

Stops Claude asking clarifying questions. Use when you want immediate action.

Prevents over-engineering. Critical when you want surgical fixes, not rewrites.

Triggers maximum reasoning budget. Use for hardest problems. Also: "think hard", "think harder", "think very hard".

Eliminates unnecessary explanation. Use when you want the answer, not the commentary.

Ensures Claude reads full context before touching anything. Avoids half-baked edits.

Assigns an expert persona. Changes tone, depth, and focus of response.

Specify the exact output format. Eliminates guessing and post-processing.

Forces structured task decomposition. Great for complex multi-file changes.

Creates git checkpoints throughout the task. Lets you review or roll back incremental changes.

Enforces test-driven development. Ensures correctness from the start.

Explain this like I'm a complete beginner. Great for unfamiliar codebases.

Triggers critical analysis. Ask Claude to critique before fixing.

Resume a cut-off response. Claude picks up exactly where it stopped.

Forces balanced analysis before a recommendation.

Claude asks questions to help you think through the problem yourself.

Full security audit. Checks for OWASP top 10, injection, auth, data exposure.

Analyse for performance bottlenecks with estimated impact per fix.