Every Claude Code slash command, ordered by how often you actually use it
The official reference lists over 110 commands. Here they are regrouped into the six you use daily, the settings ones, the bundled Skills and Workflows, and why some never appear.
Contents
The official Claude Code command reference lists more than 110 slash commands. You do not need to know them all. About six come up every day; the rest are things you look up when something goes wrong. This article regroups the official list by how often you actually reach for each one.
One thing to get out of the way first: the built-in command list has moved.
KEY POINT
What you will learn
- Where the official list lives, and how to read
<arg>versus[arg] - The six commands you use in every session, and their arguments
- What the Skill and Workflow markers mean, and why availability differs per person
Where the list lives and how to read it
The official built-in command reference is at https://code.claude.com/docs/en/commands. The /docs/en/slash-commands URL now serves a page titled "Extend Claude with skills", which has no built-in commands table. That page points you onward itself: "For built-in commands like /help and /compact, and bundled skills like /debug and /code-review, see the commands reference".
The table has just two columns, Command and Purpose. Argument notation is fixed.
| Notation | Meaning |
|---|---|
<arg> | Required. The docs say "<arg> indicates a required argument" |
[arg] | Optional. "[arg] indicates an optional one" |
From inside the CLI, /help gives you "Show help and available commands". When you want to know what you actually have rather than what exists, that is the reliable source.
The six you use every session
Learn these and day-to-day work is covered.
| Command | What the docs say (condensed) | When to reach for it |
|---|---|---|
/clear [name] | Start a new conversation with empty context. A name labels the previous conversation in the /resume picker. Aliases /reset, /new | Switching to unrelated work |
/compact [instructions] | "Free up context by summarizing the conversation so far". You can pass focus instructions for the summary | Freeing context without losing the thread |
/context [all] | Visualize context usage as a colored grid. Past the context window, the output warns how far over you are and which command frees space | Finding out what made the session heavy |
/model [model] | Switch model and save it as the default for new sessions. With no argument a picker opens; press s on a row to switch for this session only | Before and after a hard task |
/resume [session] | Resume a conversation by ID or name, or open the session picker. Alias /continue | Picking up yesterday's work |
/rewind | Rewind the conversation and/or code to a previous point, or summarize from a selected message. Aliases /checkpoint, /undo | After Claude heads the wrong way |
The split between /clear and /compact is stated right in the /clear entry: "To free up context while continuing the same conversation, use /compact instead." One discards the conversation; the other summarizes and keeps it. For the longer version see Managing context in Claude Code with /compact versus /clear.
To move the threshold where automatic compaction fires, use /autocompact [auto|<tokens>]. Pass a size such as 500k, or auto to return to the window tuned for your model; the value is saved to user settings and applied to the current session. Requires Claude Code v2.1.221 or later.
Checking settings, permissions and usage
The next most frequent group answers "what is the state of things right now".
| Command | What the docs say (condensed) |
|---|---|
/permissions | Interactive dialog for allow, ask and deny rules. View rules by scope, add or remove them, manage working directories, and review recent auto mode denials. The Auto mode tab also shows and edits classifier rules. Alias /allowed-tools |
/config [key=value ...] | Open the Settings interface, or set a value directly without opening it, such as /config theme=dark. /config --help lists the keys it accepts. Alias /settings |
/status | Open Settings on the Status tab: version, model, account and connectivity. Works while Claude is responding |
/usage | Usage. /cost is "Alias for /usage", and /stats opens it on the Stats tab |
/memory | Edit CLAUDE.md files, enable or disable auto memory, and view auto memory entries |
/hooks | View hook configurations |
/mcp | Manage MCP server connections and OAuth. Pass reconnect, enable or disable with a server name or all to change state without opening the list |
/add-dir <path> | Add a working directory for file access during this session. Partial paths suggest matches; Tab accepts one |
/cd <path> | Move the session to a new working directory, keeping the conversation |
/add-dir carries a caveat worth knowing: "Most .claude/ configuration isn't discovered from the added directory." For reading the permissions dialog, see Claude Code /permissions shows every rule and the settings file it came from.
Handing the work itself over, with Skills and Workflows
Some rows in the list carry a marker. The official definitions:
用語解説
Skill: "a bundled skill. It works like skills you write yourself: a prompt handed to Claude."
Workflow: "a bundled dynamic workflow that fans work out across many subagents and runs in the background."
Both only run on demand — the docs make the point with /verify and /deep-research: "runs only when you invoke it."
| Command | Kind | What it does |
|---|---|---|
/code-review | Skill | Review the current diff, or a PR number, branch or path you pass, for correctness bugs. --fix applies findings, --comment posts them to the PR or merge request. Alias /review |
/simplify [target] | Skill | Review changed code for cleanup opportunities and apply the fixes, with four review agents in parallel. "The review doesn't look for correctness bugs" — that is /code-review's job |
/debug [description] | Skill | Turn on debug logging for this session and troubleshoot from the log. Logging is off by default, so running it mid-session captures from that point forward |
/doctor | Skill | A setup checkup over installation, settings, extension and CLAUDE.md problems, proposing fixes you confirm. Alias /checkup |
/batch <instruction> | Skill | Parallelize a large change: decompose into 5 to 30 independent units, present a plan, then run one background subagent per unit in an isolated worktree |
/deep-research <question> | Workflow | Fan out web searches, fetch and cross-check sources, and synthesize a cited report |
/security-review | — | Analyze the changes on your current branch for security vulnerabilities. Needs an origin remote |
To build commands on the same mechanism yourself, see Create your own slash commands in Claude Code with a single Markdown file.
Commands that never appear, and ones that went away
A command can be in the list and still be missing from your session. That is by design: "Not every command appears for every user. Availability depends on your platform, plan, and environment." The docs give two examples.
/desktopshows only on macOS and x64 Windows, signed in with a Claude subscription/upgradedoes not show on Enterprise plans
Some commands are genuinely gone.
| Command | Status |
|---|---|
/pr-comments | Removed in v2.1.91. Ask Claude directly to view pull request comments |
/ultraplan | Removed. Use plan mode instead |
/agents | Now "Print a reminder to ask Claude to create or manage subagents". On v2.1.197 and earlier it opened an interactive management interface |
Some commands are only hidden from the menu
/heapdump, /setup-bedrock and /setup-vertex are documented as "Hidden from the command menu" — they do not autocomplete, but typing them in full works. Setting CLAUDE_CODE_USE_BEDROCK=1 or CLAUDE_CODE_USE_VERTEX=1 makes the matching wizard appear in the menu.
Summary
- The official built-in list is at
/docs/en/commands;/docs/en/slash-commandsis now the skills page - The table has two columns, and
<arg>is required while[arg]is optional - The daily six are
/clear,/compact,/context,/model,/resumeand/rewind - Skill means a prompt handed to Claude; Workflow means work fanned out across subagents. Both run only when invoked
- Platform, plan and environment decide what you see, so check your own list with
/help
FAQ
- Where is the official slash command list?
- At https://code.claude.com/docs/en/commands. The older /docs/en/slash-commands URL now serves the "Extend Claude with skills" page, which carries no built-in commands table. Inside the CLI, /help lists what you actually have.
- A command in the list doesn't appear in my session. Why?
- The docs state "Availability depends on your platform, plan, and environment". Their examples: /desktop shows only on macOS and x64 Windows when signed in with a Claude subscription, and /upgrade doesn't show on Enterprise plans.
- When do I use /clear instead of /compact?
- /clear starts a new conversation with empty context. To free context while staying in the same conversation, use /compact — the docs say exactly that in the /clear entry.
- What do <arg> and [arg] mean in the table?
- They are the reference's own notation: <arg> is a required argument and [arg] is an optional one.
Questions & answers
Q. My CLAUDE.md requires replies in one language, but Claude switches to English right after reading English material. Is there a reliable way to pin the reply language?
No official setting for the reply language could be found in the memory or output-styles documentation — there is no key or frontmatter field that addresses natural language at all. That said, three things about the mechanism are documented and useful here.
First, how CLAUDE.md reaches the model: "CLAUDE.md content is delivered as a user message after the system prompt, not as part of the system prompt itself. Claude reads it and tries to follow it, but there's no guarantee of strict compliance, especially for vague or conflicting instructions."
Second, there are places higher up the stack. If you want the instruction on every request, an output style sits structurally stronger.
| Mechanism | How it is delivered |
|---|---|
| CLAUDE.md | A user message after the system prompt. Re-injected from disk after compaction |
| Output style | "Claude Code sends the active style's instructions with every request." Across compaction, "System prompt and output style" both still apply |
--append-system-prompt | The docs offer this "For instructions you want at the system prompt level" |
Output styles are Markdown files in ~/.claude/output-styles or .claude/output-styles.
---
name: Portuguese only
description: Always respond in Brazilian Portuguese
keep-coding-instructions: true
---
Write every response in Brazilian Portuguese. Even when the material you read is in English, keep summaries, explanations and proposed commit messages in Portuguese. Quote code, file paths and error messages verbatim in their original language.Without keep-coding-instructions: true, a custom style drops Claude Code's built-in software engineering instructions. Switch with /output-style <style> (Claude Code v2.1.269 or later). The docs are explicit about the ceiling, though: "An output style gives Claude instructions to follow. It doesn't guarantee that something always happens or never happens."
Third, the documented ways instructions get diluted. These four are in the docs; a claim that reading English material weakens an instruction is not.
- File length: "target under 200 lines per CLAUDE.md file. Longer files consume more context and reduce adherence"
- Truncation of on-demand files and of skill bodies
- Compaction: anything given only in conversation is summarized away, while project-root CLAUDE.md is re-injected from disk
- Stale conversation crowding out context
Nothing delivered as an instruction is enforced
The docs route certainty elsewhere: "To block an action regardless of what Claude decides, use a PreToolUse hook instead." A reply's language is output content, so a hook cannot constrain it. In practice the workable setup is to carry the instruction in an output style, so it rides every request, and to delete the duplicate wording from CLAUDE.md so the two cannot conflict.
For where /output-style sits among the commands, see Every Claude Code slash command, ordered by how often you use it.
Primary sources
This article was drafted by AI from official documentation and reviewed by the site operator before publishing. Found a mistake? Let us know via the contact page.