Claude Code Commands Cheat Sheet (November 2026): Slash Commands, Shortcuts, CLI Flags and Config Files

Claude Code ships updates almost daily, so most cheat sheets online are already stale. This one is checked against Claude Code 2.1.30x (late October to early November 2026). I have grouped the commands the way you actually use them: getting set up, the slash commands you will type every day, keyboard shortcuts, CLI flags for scripts, and the config files where memory, skills, subagents, hooks and MCP servers live. Bookmark it and use Ctrl+F.

Install, Update and Sign In

# macOS / Linux
curl -fsSL https://claude.ai/install.sh | bash

# macOS with Homebrew
brew install --cask claude-code

# Windows PowerShell
irm https://claude.ai/install.ps1 | iex

# check and sign in
claude --version
claude auth login

# update
claude update

Claude Code works with a Claude Pro, Max, Team or Enterprise plan, or with API billing. If you are not sure which plan fits, see my Claude Free vs Pro vs Max vs Ultra guide; for API costs see Claude API pricing.

Session and Context Commands

Model, Effort and Config Commands

Project, Review and Planning Commands

Account, Usage and System Commands

Keyboard Shortcuts

CLI Flags and Subcommands

claude "fix the failing test"      # start with a prompt
claude -c                          # continue last session
claude -r my-feature               # resume a named session
claude -w feature/login            # work in a new git worktree
claude --model opus --effort high  # pick model and effort
claude --permission-mode plan      # start read-only

# non-interactive (scripts, CI)
claude -p "summarize changes since main" --output-format json
claude -p "..." --max-turns 10 --max-budget-usd 2 --allowedTools "Read,Grep"

# background agents
claude --bg "upgrade dependencies and run tests"
claude agents            # list running agents
claude attach <id>       # attach to one
claude logs <id>         # read its output
claude stop <id>

# MCP servers
claude mcp add --transport http github https://example.com/mcp
claude mcp add --scope user mytool -e API_KEY=xxx -- npx my-mcp
claude mcp list
claude mcp remove github

# maintenance
claude update
claude purge --dry-run
claude import codex --dry-run

Want good servers to add? See my 20 best MCP servers with install commands.

Permission Modes, Models and Effort Levels

Permission modes

• manual (config value default): asks before edits and commands.
• plan: read-only, Claude proposes a plan.
• acceptEdits: file edits go through, commands still ask.
• dontAsk: only pre-approved tools run.
• auto: a safety classifier decides; the default for new interactive sessions with no mode set.
• bypassPermissions: no checks. Use only inside containers or sandboxes.

Model aliases

Effort levels

low, medium, high, xhigh, max and auto, set with /effort, --effort or the CLAUDE_CODE_EFFORT_LEVEL environment variable. Higher effort means better answers on hard problems but uses your limits faster.

Config Files: CLAUDE.md, Settings, Skills, Agents, MCP

Skill frontmatter takes name, description, argument-hint and user-invocable. Subagent frontmatter takes name, description, tools, model and omitClaudeMd. I go deeper on skills in my Claude Skills guide.

Hook Events

Hooks run your own scripts at fixed points, configured under hooks in settings.json. Available events:

SessionStart, UserPromptSubmit, PreToolUse, PermissionRequest, PostToolUse, PostToolUseFailure, Stop, SubagentStop, PreCompact, Notification, SessionEnd, FileChanged, CwdChanged, DirectoryAdded, WorktreeCreate, PreModelSwitch.

Common uses: run a formatter in PostToolUse after every file edit, block edits to secrets in PreToolUse, or play a sound on Notification when Claude needs you.

Removed or Changed Commands

• /vim and /pr-comments were removed.
• /agents no longer runs a creation wizard. Write the agent file in .claude/agents/ yourself.
• claude project purge became claude purge.
• /cost is now an alias of /usage.

If something is missing in your install, run claude update and /release-notes; commands vary by version and plan.

Quick Start: Your First Session in 6 Steps

If you have just installed Claude Code, this is the order I’d use the commands above in a real project. It takes about ten minutes and leaves you with a setup you can reuse.

1. Install and check the version. Run the installer for your system, then claude --version. If the shell says the command is not found, see the troubleshooting table below.
2. Start inside your project. cd into the repo and run claude. Sign in with your Claude account in the browser window that opens, or use claude auth login on a remote machine.
3. Create project memory. Run /init so Claude scans the repo and writes a starter CLAUDE.md. Open it with /memory and add your build, test and lint commands.
4. Plan before editing. Press Shift+Tab until plan mode is active, describe the change, and review the plan. Switch back to an editing mode once you are happy with it.
5. Review and keep context lean. Check the work with /diff, and use /compact with a focus when the session gets long, or /clear when you move to an unrelated task.
6. Check your limits and resume later. Run /usage to see how much of your plan you used, name the session with /rename, and pick it up next time with claude -c or /resume.

Troubleshooting Common Claude Code Problems

When something breaks, start with /doctor inside a session, or claude doctor from your shell if Claude Code will not start at all. These are the problems I see asked about most, with the fixes from Anthropic’s troubleshooting pages.

If settings, hooks or skills do not seem to apply, remember that ~/.claude.json holds app state, while permissions, hooks and env belong in ~/.claude/settings.json. Mixing those two files up is one of the most common configuration mistakes.

Frequently Asked Questions

How do I see all Claude Code commands?

Type / in the prompt to get the live list for your version, or /help. Press ? in the transcript viewer for shortcuts.

What is the difference between /clear and /compact?

/clear drops the conversation entirely. /compact replaces it with a summary, so Claude keeps the key decisions while freeing context.

How do I check my Claude Code usage limits?

Run /usage (or /cost). If you keep hitting limits or errors, see my Claude 429 and 529 error fix guide.

How do I stop Claude Code asking for permission every time?

Use Shift+Tab to switch to accept-edits or auto mode, add allow rules with /permissions, or run /fewer-permission-prompts to build an allowlist from your history.

Claude Code vs Codex vs Cursor: which should I use?

See my coding agent comparison and OpenAI Codex guide.

Why does my terminal say claude: command not found?

The install worked but the folder holding the binary, ~/.local/bin on macOS and Linux, is not on your PATH. Add it to your shell profile, reload the shell, and run claude --version again. If you only installed the VS Code extension, run the standalone installer too, because the extension keeps its own private copy.

How do I fix an MCP server that will not connect?

Run /mcp to see its status. Project servers from .mcp.json need a one-time approval there, failed servers often use relative paths, and a connected server with zero tools usually needs a Reconnect. For details, start with claude --debug=mcp.

Sources

• ComputingForGeeks: Claude Code cheat sheet (checked against 2.1.304, October 26, 2026)
• Anthropic: Claude Code documentation
• GitHub: Claude Code slash commands cheatsheet