Claude Code slash commands: the full list and a cheat sheet

Claude Code has 112 built-in slash commands as of October 7, 2026, according to Anthropic's command reference; type / in a session to see the ones available to you. The ones you'll use most: /clear to start fresh, /compact to summarize a long conversation, /model and /effort to choose the model and how hard it thinks, /plan to plan before editing, /context and /usage to see what's using tokens, /init to write a CLAUDE.md, /memory to edit it, /code-review to review changes, /mcp and /plugin to manage tools, and /doctor to fix setup problems. Custom commands are now skills: a file in .claude/commands or .claude/skills becomes a /command, with $ARGUMENTS for what you type after it.

Key takeaways

• Type / for the list; Tab completes, and /help explains.
• Aliases: /clear is also /new and /reset; /usage is also /cost and /stats; /review runs /code-review.
• Custom commands: a Markdown file in .claude/commands/ or a skill in .claude/skills/; $0 is the first argument.
• Gone: /ultraplan, /vim, /pr-comments and the # memory shortcut.
• Keys worth knowing: Shift+Tab for modes, Esc to stop, Esc twice to rewind, Ctrl+G to edit the prompt or plan in your editor.

What are the most useful Claude Code slash commands?

From Anthropic's command reference (Anthropic), grouped:

Session

• /clear (/new, /reset): start a new conversation.
• /compact [instructions]: summarize the conversation to free context.
• /resume (/continue): reopen an earlier session.
• /rewind: roll back code, conversation or both.
• /btw: ask a side question that stays out of the conversation.
• /branch: branch the conversation; /fork copies it into a background session.
• /goal [condition]: keep working until a condition is met.
• /export, /copy, /rename, /diff and /exit.

Context and cost

• /context: what's filling the context window (Claude Code context window).
• /usage (/cost, /stats): cost, plan limits and what's using them.
• /autocompact: when to compact automatically.
• /skill-doctor: what each skill costs in context.

Models and modes

• /model, /effort and /fast: model, reasoning effort and fast mode.
• /plan [task]: plan before editing (Claude Code plan mode).
• /permissions, /sandbox and /config.

Project setup

• /init: write a starter CLAUDE.md.
• /memory: edit CLAUDE.md and auto memory (Claude Code memory).
• /import codex (or gemini, cursor): bring another agent's setup over.
• /skills and /hooks: see what's loaded.

Review and git

• /code-review [low to max, or ultra] (/review): review changes or a pull request (Claude Code GitHub Actions).
• /security-review: check your branch for security problems.
• /simplify: clean-up suggestions.
• /batch: split a large change across parallel subagents.
• /install-github-app: set up Claude in GitHub.

Integrations

• /mcp and /plugin: tools and plugins (Claude Code MCP, Claude Code plugins).
• /ide and /chrome: connect your editor or Chrome.
• /remote-control (/rc), /mobile and /desktop: continue elsewhere (Claude Code on your phone).
• /schedule: routines on a schedule.

Account and diagnostics

• /login, /logout, /usage-credits and /upgrade.
• /doctor: check and fix your setup; /doctor prompt-audit checks instruction files.
• /bug and /feedback: report a problem.

Some commands depend on your platform and plan.

How do you make a custom slash command?

Custom commands have been merged into skills (Anthropic). Create .claude/commands/fix-issue.md in a project, or ~/.claude/commands/ for yourself:

---
description: Fix a GitHub issue by number
argument-hint: [issue-number]
---
Fix GitHub issue #$ARGUMENTS: read it with gh issue view $0,
implement the fix, and add tests.

Then run /fix-issue 123. The same file as .claude/skills/fix-issue/SKILL.md works too and can bring supporting files.

• Arguments: $ARGUMENTS is everything you typed; $0 is the first argument and $1 the second (counting from zero, unlike many older guides); named arguments come from an arguments: list in the frontmatter.
• Shell output: a line starting with ! and a backtick-quoted command runs before Claude sees the prompt, so the output is included.
• Folders become namespaces: .claude/commands/frontend/component.md is /frontend:component.
• Plugins and MCP: plugin commands run as /plugin-name:command; MCP prompts as /mcp__server__prompt.
• Who can run it: Claude can run a skill itself unless it has disable-model-invocation: true.

Claude Code keyboard shortcuts

From Anthropic's interactive-mode reference (Anthropic):

• Shift+Tab: cycle permission modes (Alt+M on some Windows terminals).
• Esc: stop Claude and keep its work. Esc Esc on an empty prompt: rewind.
• Ctrl+C: interrupt or clear; twice on an empty prompt to exit.
• Ctrl+G: open the prompt, or a proposed plan, in your editor.
• Ctrl+R: search your prompt history.
• Alt+P: switch model. Alt+O: toggle fast mode.
• ! at the start: run a shell command. @: reference a file.
• Ctrl+V: paste an image (Alt+V on Windows).
• Shift+Enter or \+Enter: a new line.

What changed in 2026?

New this year: /plan, /effort, /loop, /goal, /recap, /simplify and /batch, /cd, /subtask, /skill-doctor and /doctor prompt-audit. Removed: /vim (April), the /agents wizard (July) and /ultraplan (August); /review became an alias of /code-review in August, and the # shortcut for memory went in December 2025 (Anthropic).

Codex slash commands

Codex's CLI has its own set (OpenAI): /init writes an AGENTS.md, /plan, /review, /compact, /model, /status, /mcp, /skills, /memories, /resume, and /import to bring in a Claude Code setup. More in How to use Codex.

Commands in a shared room

In Poly (usepoly.co), a team works with one Claude Code or Codex agent in a shared browser room using the repository's own commands and skills, everyone sees each step live, and any member can approve a change before it happens. Free to start. What is Poly?