Claude Code Cheat Sheet (2026) — Commands & Shortcuts

The ten commands below cover about 90% of a Claude Code session: /clear to reset context, /compact to summarize it, /context to see what is eating it, /model to switch models, /rewind to undo code and conversation, Shift+Tab to cycle permission modes, Esc to interrupt, Esc Esc to rewind, @ to attach a file, and ! to run a shell command inline. Everything else in this cheat sheet is the long tail you reach for less often but want on hand.

This is current for 2026 (Claude Code v2.1.x) and sourced from the official command reference, CLI reference, and interactive-mode docs. Availability varies by platform, plan, and version — check /help in your own install for the authoritative list.

Slash commands — the daily drivers

Type / at the start of a message to filter these. The ones you will actually wear out:

Command	What it does
/clear (aliases /new, /reset)	Start a fresh conversation with empty context
/compact [instructions]	Summarize the conversation to free up context without losing the thread
/context [all]	Visualize current context usage as a colored grid — find what is bloating it
/model [model]	Switch model mid-session
/rewind	Roll code and conversation back to a checkpoint
/plan	Switch into plan mode before a large change
/effort [level]	Set model effort: low, medium, high, xhigh, max, ultracode
/usage (alias /cost)	Show token usage and cost for the session
/resume [id]	Return to an earlier conversation or session
/config (alias /settings)	Open settings — theme, model, preferences
/memory	Edit CLAUDE.md memory files and manage auto-memory
/init	Generate a CLAUDE.md guide for the current project

Slash commands — review, workflow, and power tools

Command	What it does
/code-review [level] [--fix] [pr# | branch | path] (alias /review)	Review a diff or PR for correctness bugs and cleanup
/security-review	Check the diff for security vulnerabilities
/pr-comments [--fix]	Post review findings as inline GitHub PR comments
/verify	Verify that code changes actually work
/simplify	Remove duplication and complexity from changed code
/agents	Manage subagent configurations
/subtask [prompt]	Hand a side task to a subagent; its result returns to your conversation
/background [prompt] / /fork [prompt]	Detach or copy the session to run as a background agent
/loop [interval] [prompt]	Run a prompt repeatedly while the session stays open
/goal [condition]	Set a goal — Claude keeps working until the condition is met
/batch	Orchestrate large-scale changes across a codebase in parallel
/hooks	View hook configurations for tool events
/permissions	Set approval rules for file and command execution
/mcp	Manage MCP server connections and OAuth
/diff	Open an interactive diff of uncommitted changes
/export [file] / /copy [N]	Export the conversation as text / copy the last response
/doctor	Run a setup checkup that diagnoses and can fix issues
/vim	Open the response in your $EDITOR

One structural change worth knowing: custom slash commands are now skills. A file at .claude/commands/deploy.md and a skill at .claude/skills/deploy/SKILL.md both create /deploy. Existing .claude/commands/ files keep working; skills add a supporting-file directory and frontmatter to control whether you or Claude invokes them.

CLI commands

What you run in the terminal before the session starts:

Command	What it does
claude	Start an interactive session
claude "query"	Start interactive with an initial prompt
claude -p "query"	Print mode — query via SDK, then exit (great for scripts and pipes)
cat file | claude -p "query"	Pipe content in and process it headless
claude -c	Continue the most recent conversation in this directory
claude -r "" "query"	Resume a session by ID or name
claude mcp	Configure MCP servers
claude agents	Open the agent view to monitor and dispatch parallel background sessions
claude doctor	Print installation and settings diagnostics
claude update / claude install	Update to latest / install the native binary
claude setup-token	Generate a long-lived OAuth token for CI and scripts

CLI flags worth knowing

Flag	What it does
--print, -p	Print the response without interactive mode
--continue, -c	Load the most recent conversation in the current directory
--resume, -r	Resume a specific session by ID or name
--model	Set the model (alias or full name)
--permission-mode	Begin in a specific permission mode (e.g. plan, acceptEdits, auto)
--dangerously-skip-permissions	Skip permission prompts entirely (bypass mode — use with care)
--add-dir	Add extra working directories Claude can read and edit
--allowedTools / --disallowedTools	Allow tools without prompting / deny specific tools
--agents	Define custom subagents dynamically via JSON
--max-budget-usd	Cap the dollar spend on API calls before stopping
--max-turns	Limit the number of agentic turns
--effort	Set the effort level for the session
--fallback-model	Auto-fall back to another model when the primary is unavailable
--worktree, -w	Create and work in a git worktree
--output-format	Output format for print mode: text, json, stream-json
--verbose, -v	Verbose output

Keyboard shortcuts

Key	What it does
Esc	Interrupt Claude mid-turn (keeps work so far), or close an open dialog
Esc Esc	Clear the input draft; on an empty prompt, open the rewind menu
Shift+Tab	Cycle permission modes: default → acceptEdits → plan → (auto / bypass)
Ctrl+C	Interrupt a running op; if idle, first press clears input, second exits
Ctrl+D	Exit the session (double-press to confirm)
Ctrl+L	Redraw the screen (fix a garbled terminal); twice = /clear
Ctrl+R	Reverse-search command history
Ctrl+O	Toggle the transcript viewer (detailed tool usage)
Ctrl+B	Background running Bash tasks and agents
Ctrl+T	Toggle Claude's task checklist
Ctrl+V / Cmd+V	Paste an image from the clipboard as an [Image #N] chip
Option/Alt+P	Switch model without clearing your prompt
Option/Alt+T	Toggle extended thinking
Shift+Enter / Ctrl+J / \+Enter	Insert a newline (multiline input)
@	Mention a file to attach it to the prompt
!	Run a shell command inline (bash mode)
#	Add a line to CLAUDE.md memory

The three that change how a session feels

Most people learn the commands and stop there. Three habits matter more than the rest.

Watch your context with /context, not by vibes. A bloated context window makes responses slower and worse. The colored grid shows exactly what is consuming tokens — old file reads, a long tool result, a verbose subagent — so you know whether to /compact or /clear. If you want to model the token cost before you run something big, our free AI token estimator gives you the number up front. The deeper mechanics of what actually costs tokens are in our breakdown of the Claude Code cost system.

Live in Shift+Tab. Cycling permission modes is the single biggest control on how autonomous a session feels — plan mode to think before touching code, acceptEdits to stop confirming every write, back to default when you want the brakes on. It is faster than reaching for a flag on restart.

Treat CLAUDE.md as the real interface. The # shortcut and /memory let you encode project rules once instead of re-explaining them every session. A well-built CLAUDE.md plus custom skills is what separates a tidy Claude Code setup from a chaotic one — it is the same discipline packaged in our Claude Code Production Pack, which ships the agents, skills, and rule files that run a real production repo. If you are wiring in your own tools, the MCP server build guide walks the TypeScript path.

Keep this handy

Bookmark this page and check /help in your install for anything version-specific — the surface moves fast, and this reflects Claude Code v2.1.x in 2026. The commands are free; the setup that makes them pay off is the part worth building.

Every Claude Code product mentioned is available at wowhow.cloud — pay once, ship forever.