Claude Code Slash Commands: The Complete List

The list that pops up when you hit / in Claude Code is intimidating at first. A hundred-odd commands, half of them named in a way that gives nothing away, and your own custom commands mixed in with the built-ins.

You don’t need all of them. Ten commands cover almost everything you do day to day, and the rest is reference material you look up when the moment arrives. This page gives you both: the ten first, then the full list grouped by what it’s for 🧰

The Short Answer

How Slash Commands Work

A slash command isn’t a regular message to Claude. It’s an instruction to Claude Code itself, or a prepackaged prompt. A few rules govern them:

The command has to start the message. A /compact in the middle of a sentence is just text, not a command. Everything you type after the command name becomes its arguments.

Sending one mid-response queues it. If Claude is working when you send a command, it runs after the current turn finishes. /status, /tasks and /usage are the exceptions: they run immediately without interrupting anything.

Skills chain. Bundled skills and your own can be stacked: /skill-a /skill-b do the thing loads every skill named at the start and passes the trailing text to each one as arguments. Up to six at a time.

The 10 You Actually Use

Learn these and you’ll rarely need to look up the rest.

Full List: Session and Conversation

All of context management lives here, and these are the commands that save you the most time.

Model, Effort and Interface

Code, Git and Review

Project, Context and Memory

Configuration and Diagnostics

Commands that depend on platform, plan or environment:

Bundled Skills and Workflows

These aren’t classic commands, they’re prepackaged prompts handed to Claude. You can invoke them, and Claude can load some of them on its own when they’re relevant.

Commands That No Longer Work

Still quoted in older guides and blog posts, but gone:

Writing Your Own Slash Command

Custom commands and skills are the same thing now. .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md both produce /deploy and behave identically.

The Easy Way: Have Claude Write It

You don’t have to open the file yourself. Describe what you want and let Claude Code create the skill:

Create a slash command called /deploy for me.

It takes an environment name as its argument (staging, production).
It should, in order:
1. Run npm test and stop if anything fails
2. Build with npm run build
3. Deploy to the given environment

Make it project-scoped, write it under .claude/skills/.
It has side effects, so it shouldn't fire on its own.

Those last two lines matter. “Project-scoped” puts it under .claude/skills/, “personal” puts it under ~/.claude/skills/. “Shouldn’t fire on its own” adds disable-model-invocation: true to the frontmatter.

Once the file exists, run /reload-skills and it shows up without a restart.

If You’d Rather Write It Yourself

The quickest route is a single file:

---
name: deploy
description: Deploys the app to the given environment
disable-model-invocation: true
---

Deploy to $ARGUMENTS:

1. Run the test suite
2. Build the app
3. Push to the deploy target

Type /deploy staging and $ARGUMENTS becomes staging.

The command name comes from the folder name. Anything with side effects (/deploy, /commit, sending messages) should carry disable-model-invocation: true, otherwise Claude can fire it whenever it decides the moment is right.

For how skills, subagents and hooks differ, the full set of frontmatter fields and hook examples, see the skills, subagents and hooks guide .

MCP Server Commands

Connected MCP servers can expose slash commands too. They show up in this shape:

/mcp__<server-name>__<prompt-name>

A GitHub MCP server’s list_prs prompt becomes /mcp__github__list_prs. If the server isn’t connected the command won’t appear, so check /mcp for connection status.

The Terminal Side: claude Commands

Slash commands run inside a session. These start one:

The flags worth knowing:

Frequently Asked Questions (FAQ)

How many slash commands does Claude Code have? There are over a hundred built-in commands and the count changes with each release. On top of that come your own skills, any plugins you installed and prompts exposed by connected MCP servers. Run /help to see the current list for your own setup.

What should I do when Claude Code runs out of context? Run /context first to see what filled the window, then /compact to summarize the conversation. /compact keeps the same conversation going, so Claude still remembers what it did and why. If the topic has changed entirely, /clear is the better choice.

What is the difference between /clear and /compact? /compact reduces the history to a summary but continues the same conversation. /clear ends the conversation and empties context completely. Use /clear mid-task and you will have to explain the situation from scratch.

How do I undo a change Claude Code made? /rewind rolls back both code and conversation to a checkpoint. Pass a number to rewind that many turns, pass text to search for a message, or run it with no argument to open a menu of recent checkpoints.

How do I create my own slash command in Claude Code? Create .claude/skills/<name>/SKILL.md in your project; the command name comes from the folder name. A single .claude/commands/<name>.md file does the same job. Use $ARGUMENTS inside it and whatever you type after the command gets substituted in.

What is the difference between slash commands and skills in Claude Code? There is no longer a difference. Custom slash commands were merged into skills, so .claude/commands/deploy.md and .claude/skills/deploy/SKILL.md both create /deploy. What skills add is a folder for supporting files and the ability for Claude to load them automatically when relevant.

How do I check token usage in Claude Code? /usage shows token usage and cost for the session, and /cost is an alias for the same command. It runs immediately without interrupting the response. For a detailed breakdown of what is filling the context window, use /context.

Does a slash command work in the middle of a sentence? No. A command is only recognized at the start of a message; anywhere else it is treated as plain text. Everything typed after the command name becomes its arguments.

A command isn’t showing up in my menu, what now? Three possible reasons. If your version doesn’t have it yet, run claude update. The command may depend on platform, plan or environment: /upgrade is hidden on Enterprise plans, /desktop only appears on macOS and x64 Windows, and /setup-bedrock only when the matching environment variable is set. The third possibility is that it was removed: /vim, /pr-comments and /ultraplan are gone. Run /help for the current list on your own install.

Wrapping Up

The command list is long but it revolves around one thing: context management. /context, /compact, /clear and /rewind are the backbone of daily use, and most of the rest gets opened once a month.

The second ring is /model, /effort, /diff and /code-review. Once they click, they’re also daily.

For the other hundred, bookmark this page instead of memorizing it. The only one worth memorizing is /help 😄

If you want to shape these commands around your own workflow, the skills, subagents and hooks guide is the next step, and the MCP guide covers adding tools 👀