The Claude Code cheat sheet that's actually current

Add a working directory for file access during the current session.

Enable or disable the advisor tool, which consults a second model for guidance.

Reminder to ask Claude to create or manage subagents, or edit agent configurations.

Spawn a Claude Code on the web session that watches the current branch's PR and pushes fixes.

Detach the current session to run as a background agent and free up the terminal.

Orchestrate large-scale changes across a codebase in parallel using worktrees.

Create a branch of the current conversation to try a different direction.

Ask a quick side question without adding it to the conversation history.

Report a bug, optionally sharing your conversation with session history.

Move this session to a new working directory.

Configure the Claude in Chrome browser integration.

Load Claude API reference material and Managed Agents reference into context.

Start a new conversation with empty context. Resets input history for the session.

Review the current diff for correctness bugs and cleanup opportunities.

Set the prompt bar color for the current session.

Free up context by summarizing the conversation so far.

Open the Settings interface, or set preferences directly from the command line.

Visualize current context window usage as a colored grid.

Copy the last assistant response (or the Nth one back) to the clipboard.

Alias for /usage - show token usage and cost for the current session.

Get design guidance for charts, graphs, and dashboards before writing chart code.

Enable debug logging and troubleshoot a runtime issue.

Fan out web searches on a question and synthesize a cited report.

Authorize design-system access so /design-sync can run.

Convert your repo's React design system and upload it to Claude Design.

Continue the current session in the Claude Code Desktop app.

Open an interactive diff viewer showing uncommitted changes.

Run a setup checkup that diagnoses and fixes installation issues.

Set the model's reasoning effort level for the session.

Exit the CLI.

Export the current conversation as plain text.

Toggle fast mode on or off for the session.

Send product feedback about Claude Code directly to Anthropic.

Scan your transcripts and add an allowlist to reduce permission prompts.

Toggle a focus view that shows only essential information.

Copy the current conversation into a new background session.

Set a goal that Claude keeps working toward across multiple turns.

Write a JavaScript heap snapshot for diagnosing memory usage.

Show help and the full list of available commands.

View the current hook configurations for tool events.

Manage IDE integrations and show connection status.

Initialize the project with a generated CLAUDE.md guide.

Generate a report analyzing your Claude Code sessions.

Install the Claude GitHub App for a repository.

Install the Claude Slack app (Claude Tag) for a workspace.

Open your keyboard shortcuts file for editing.

Sign in to your Anthropic account.

Sign out of your Anthropic account.

Run a prompt repeatedly on an interval while the session stays open.

Manage MCP server connections and OAuth authentication.

Edit CLAUDE.md memory files and manage auto-memory.

Show a QR code to download the Claude mobile app.

Switch the active model and save the choice as your default.

Share a free week of Claude Code with friends.

Set approval rules for file access, tool use, and other sensitive operations.

Manage Claude Code plugins: install, enable, disable, update.

Post code review findings as inline comments on a GitHub PR.

Summarize the conversation so far in one message.

Rename the current session.

Return to an earlier conversation or branch.

Give a fast, single-pass, read-only review of a GitHub pull request.

Roll code and conversation back to a checkpoint, or summarize part of the conversation.

Also reachable by pressing Esc twice on an empty prompt.

Check the current diff for security vulnerabilities.

Show current session status and background task progress.

Configure the status bar shown at the bottom of the terminal.

List the current session's background work, including subagents that have finished.

Configure terminal integration, including the Shift+Enter multiline binding.

Set the display theme.

Manage a to-do list for the current session.

Toggle between fullscreen and split-pane rendering modes.

Check for and install Claude Code updates.

Upgrade your Claude subscription plan.

Show token usage and costs for the current session.

Skill that verifies code correctness and test coverage.

Toggle vim-style keybindings for the prompt input.

Enable or disable voice input and output.

View and manage dynamic workflows.

Interrupt a running operation. On empty input, one press clears the prompt and a second exits Claude Code.

Stop all running background subagents in this session.

Press twice within 3 seconds to confirm.

Exit the Claude Code session (or delete the character after the cursor when the prompt has text).

Open your prompt in your default text editor.

Redraw the screen if the display looks garbled.

Press twice within two seconds to also run /clear.

Toggle the transcript viewer, showing detailed tool usage with timestamps.

Reverse search through your prompt history.

Paste an image from the clipboard as an [Image #N] chip.

Move a running Bash command or agent to the background.

tmux users press it twice, since tmux claims Ctrl+B as its own prefix key.

Toggle Claude's to-do task checklist in the status area.

Stash the current prompt text, or restore it when pressed again on an empty prompt.

Suspend Claude Code to your shell (Unix only). Run fg to resume.

Cycle through tabs in permission dialogs and menus.

Move the cursor within a multiline prompt, then step through command history.

Interrupt Claude mid-turn, or close an open dialog.

Clear the input draft, or open the rewind menu when the prompt is already empty.

Cycle through permission modes: default, acceptEdits, plan, and any you've enabled.

Switch models without clearing your current prompt.

Toggle extended thinking mode.

Toggle fast mode.

Move the cursor to the start of the current line.

Move the cursor to the end of the current line.

Delete to the end of the line, storing the text for pasting.

Delete from the cursor to the start of the line, storing the text for pasting.

Delete the previous word, storing the text for pasting.

Paste text deleted with Ctrl+K, Ctrl+U, or Ctrl+W.

Cycle through previously deleted text.

Move the cursor back one word.

Move the cursor forward one word.

Undo the last input edit.

Insert a newline in the prompt. Works in every terminal with no setup.

Insert a newline after enabling Option as Meta on macOS.

Insert a newline natively in iTerm2, WezTerm, Ghostty, Kitty, Warp, Terminal.app, and Windows Terminal.

In VS Code, Cursor, Alacritty, or Zed, run /terminal-setup first to install the binding.

Insert a newline. Works in any terminal without configuration.

Paste multiline content like code blocks or logs straight into the prompt.

Open the command or skill menu.

Enter shell mode and run a command directly, without Claude's approval.

Trigger file path autocomplete.

Type a full emoji shortcode to insert it, or two-plus letters for suggestions.

Toggle the shortcut help panel.

Toggle the keyboard shortcut help panel inside the transcript viewer.

Jump to the previous or next user prompt, like a vim paragraph motion.

Write the full conversation to your terminal's native scrollback so you can search it.

Exit the transcript viewer.

Voice dictation. Hold to record, or run /voice tap for tap-to-toggle.

Turn on vim-style editing for the prompt input.

Map a two-key INSERT-mode sequence like "jj" to Escape.

Only read from user or managed settings, never a project's .claude/settings.json - a checked-out repo can't remap your keystrokes.

Enter NORMAL mode from INSERT or VISUAL mode.

Insert text before the cursor.

Insert text at the beginning of the line.

Insert text after the cursor.

Insert text at the end of the line.

Open a new line below and enter INSERT mode.

Open a new line above and enter INSERT mode.

Start character-wise visual selection.

Start line-wise visual selection.

Move left / down / up / right.

Move right.

Jump to the next word.

Jump to the end of the current word.

Jump to the previous word.

Jump to the beginning of the line.

Jump to the end of the line.

Jump to the first non-blank character on the line.

Jump to the beginning of the input.

Jump to the end of the input.

Jump to the next occurrence of a character.

Jump to the previous occurrence of a character.

Jump to just before the next occurrence of a character.

Jump to just after the previous occurrence of a character.

Repeat the last f/F/t/T motion.

Repeat the last f/F/t/T motion in reverse.

Open reverse history search, same as Ctrl+R.

Delete the character under the cursor.

Delete the current line.

Delete to the end of the line.

Delete the word, to its end, or back a word.

Change (clear and re-insert) the current line.

Change to the end of the line.

Change the word, to its end, or back a word.

Substitute the character under the cursor and enter INSERT mode.

Substitute the whole line and enter INSERT mode.

Yank (copy) the current line.

Yank the word, to its end, or back a word.

Paste after the cursor.

Paste before the cursor.

Indent the current line.

Dedent the current line.

Join the current line with the next.

Undo the last change.

Repeat the last change.

Inner or around a word (a text object, used with d, c, or y).

Inner or around a WORD, delimited by whitespace only.

Inner or around double quotes.

Inner or around single quotes.

Inner or around parentheses.

Inner or around brackets.

Inner or around braces.

Delete the selection.

Yank the selection.

Change the selection.

Replace the selection with register contents.

Replace every selected character with {char}.

Toggle, lowercase, or uppercase the selection.

Indent or dedent the selected lines.

Join the selected lines.

Swap the cursor and the anchor of the selection.

Toggle between character-wise and line-wise selection, or exit visual mode.

Requires your approval for each action.

Shown as "Manual" in the mode indicator.

Automatically accepts file edits without prompting for approval.

Requires you to review and approve Claude's plan before it executes anything.

Classifier-based mode that automatically blocks or allows actions based on configured rules.

Skips every permission check.

Recommended only in sandboxes with no internet access.

Cycles through default, acceptEdits, plan, and any other modes you've enabled.

Highest precedence. Deployed by IT to every organization member or machine user.

Project-specific and personal. Gitignored, applies to you alone in this repository.

Project scope. Committed to the repo so every collaborator gets the same rules.

User scope. Applies to every project on your machine and is never shared.

Project-scoped MCP server config, meant to be committed so the whole team gets the same servers.

User-scope MCP servers and CLI state, tracked per project, never committed.

Fires when a session begins or resumes.

Fires when you start Claude Code with --init-only, or with --init or --maintenance in -p mode.

Fires when you submit a prompt, before Claude processes it.

Fires when a user-typed command expands into a prompt, before it reaches Claude.

Fires before a tool call executes.

Fires when a tool call needs a permission decision.

Fires when a tool call is denied by the auto mode classifier.

Fires after a tool call succeeds.

Fires after a tool call fails.

Fires after a full batch of parallel tool calls resolves, before the next model call.

Fires when Claude Code sends a notification.

Fires while assistant message text is displayed.

Fires when a subagent is spawned.

Fires when a subagent finishes.

Fires when a task is being created via TaskCreate.

Fires when a task is being marked as completed.

Fires when Claude finishes responding.

One of several events that ignores matchers entirely.

Fires when the turn ends due to an API error.

Fires when an agent team teammate is about to go idle.

Fires when a CLAUDE.md or .claude/rules/*.md file is loaded into context.

Fires when a configuration file changes during a session.

Fires when the working directory changes.

Fires when a working directory is added mid-session.

Fires when a watched file changes on disk.

Fires when a worktree is being created.

The one event where any non-zero exit code aborts the action, not just exit 2.

Fires when a worktree is being removed.

Fires before context compaction.

Fires after context compaction completes.

Fires when an MCP server requests user input during a tool call.

Fires after a user responds to an MCP elicitation.

Fires when a session terminates.

Success. Claude Code parses stdout for JSON output fields - JSON is only processed on exit 0.

Blocking error. Stderr is fed back to Claude as the reason; the exact effect depends on the event (PreToolUse blocks the tool call, UserPromptSubmit rejects the prompt).

Non-blocking error for most events - the transcript shows the first line of stderr and execution continues.

Allow tool access to additional directories beyond the current working one.

Set the agent for the current session, overriding the agent setting.

Define custom agents inline as a JSON object for this session only.

Allow specific tools without a permission prompt, e.g. "Bash(git *) Edit".

Deny specific tools for this session.

Append extra instructions to the default system prompt.

Continue the most recent conversation in the current directory.

Resume a conversation by session ID, or open the interactive picker.

Bypass all permission checks.

Recommended only for sandboxes with no internet access.

Enable debug mode, optionally filtered by category, e.g. "api,hooks".

Set the reasoning effort for the session: low, medium, high, xhigh, or max.

Automatically fall back to another model when the default is overloaded.

Only works with --print.

Automatically connect to the IDE on startup if exactly one valid IDE is available.

Load MCP servers from JSON files or strings for this session.

Only use MCP servers passed via --mcp-config, ignoring every other configured source.

Set the model for the session, by alias (fable, opus, sonnet) or full name.

Set a display name for the session, shown in the prompt box and the /resume picker.

Disable the Claude in Chrome integration for this session.

Set the permission mode at startup: acceptEdits, auto, bypassPermissions, manual, dontAsk, or plan.

Load a plugin from a local directory or .zip for this session only.

Print the response and exit, for piping into other commands.

Set the print-mode output format: text, json, or stream-json.

Set the print-mode input format: text or stream-json.

Constrain print-mode output to a JSON Schema.

Cap dollar spend on API calls for this run.

Only works with --print.

Use a specific session ID for the conversation.

Choose which settings sources load: user, project, local.

Load extra settings from a file path or a JSON string.

Replace the default system prompt entirely.

Restrict the built-in tool list available this session.

Override the verbose mode setting from config.

Print the installed Claude Code version.

Create a new git worktree for this session.

Minimal mode: skip hooks, plugins, auto-memory, and CLAUDE.md auto-discovery.

Disable all customizations (CLAUDE.md, skills, plugins, hooks, MCP, themes) to troubleshoot a broken config.

Add an MCP server, local (stdio) or remote (HTTP), to Claude Code.

List every configured MCP server and its connection health.

Remove a configured MCP server.

Manage background agents from the terminal: list, dispatch, inspect.

Sign in to your Anthropic account from the CLI.

Show the current authentication status.

Check the health of your Claude Code installation from outside a session.

Install a plugin from an available marketplace.

Check for updates and install the latest version.

Delete all Claude Code state for a project: transcripts, tasks, file history, config entry.