Claude Code commands are short instructions that start with a slash and control Claude Code inside a session, for example /model to switch models or /resume to return to an earlier conversation. Type / into the input field and you immediately see the commands available to you. To get started, 15 commands are enough, and I show them here with a table and a short explanation each.
Claude Code is Anthropic's tool that lets Claude work directly in your files, in the terminal, in the desktop app or in your editor. I work with it every day. My fourteen AI employees run on Claude Team, some also on open models or through the API. Everything about the commands comes from Anthropic's official command reference, retrieved on October 26, 2026. If you have not set up Claude Code yet, start with Claude Code: A Guide for Beginners. All AI basics are on my AI Basics page.
What a command is in Claude Code
A session is an ongoing conversation with Claude Code in a folder. In it, you write normal messages and Claude works. A command is something else: it controls Claude Code itself, such as which model runs or whether the conversation starts over. Because commands begin with /, they are also called slash commands.
Three rules from the docs help you read any command list:
- A command only counts at the start of your message. Whatever follows is passed to the command as an argument, for example
/rename quote-miller.
- If you send a command while Claude is responding, Claude Code queues it and runs it after the current answer. A few commands, such as
/usage, run immediately.
- Not every command appears for everyone. According to Anthropic, it depends on your platform, plan and environment.
Getting started: /help, /init and /memory
/help is the first command to remember. It shows what exists.
/init creates a starter CLAUDE.md in a new project. That is a text file with instructions that Claude Code reads at every start, your house rules for this project. With /memory you edit it later and see what Claude has noted for itself with auto memory. What good first house rules look like is in the section "CLAUDE.md as your first house rule" of my guide for beginners. How CLAUDE.md and AGENTS.md get along is explained in CLAUDE.md vs AGENTS.md: What Survives.
While you work: /model, /plan and /permissions
/model switches the model. Without an argument, it opens a picker. Important for beginners: according to the docs, the switch is saved as your default for new sessions. If you only want to switch for the current session, press s on a row in the picker. Which models exist is in Claude Models: Opus, Sonnet, Haiku, Fable.
/plan switches into plan mode before a larger change. In plan mode, Claude reads files and writes a plan but doesn't edit your files. If you add a description right away, such as /plan sort the customer list by postcode, Claude starts on that task directly. /permissions opens the rules for what Claude may do without asking. What permissions and the sandbox mean in detail is in Claude Code Security: Permissions and Sandbox.
When the conversation gets long: /context, /compact and /clear
Everything Claude has in front of it in a session is called context. It is limited. What that means is explained tool-neutrally in Context and Context Windows in AI Explained.
/context shows what is taking up space and suggests where to save. With /context all you see every item./compact summarises the conversation so far and frees up space without ending the conversation. You can add a focus so the important parts stay in the summary./clear starts a new conversation with empty context. According to the docs, the project's memory stays in place. The old conversation is not gone: you bring it back with /resume.
Between sessions: /resume, /rename and /rewind
/resume opens a picker with your earlier conversations. To find things there, give sessions a name with /rename, for example /rename invoice-template. /clear also takes a name, and the old conversation then appears under it in the picker.
/rewind is the reverse gear. You roll back the conversation, the code or both to an earlier point. According to the docs, you can also use it to summarise only part of the conversation from a selected message.
When something is off: /diff, /usage and /doctor
My most common mistake at the start was to wave changes through without reading them. That is what /diff is for: it shows what has changed in your folder, including what Claude edited.
/usage shows what the session has used and where you stand against your plan's limits. /doctor checks your setup, finds problems in installation, settings and CLAUDE.md, and proposes fixes that Claude only applies after you confirm.
Commands for extensions
As soon as you extend Claude Code, more commands come into play. They are not among the first 15, but you will meet them quickly:
According to the docs, you create your own commands as skills. How skills, subagents and agents fit together is in Claude Subagents, Skills, Agents. Which extension is for what is sorted out in Claude Code Extensions: Plugins, Mods, Skills.
Four keyboard shortcuts worth knowing
Besides the commands, four keys from the interactive mode docs help you:
Esc stops the current answer so you can redirect. What Claude has done so far stays.Esc twice clears your draft if there is text in the input field. With an empty field, it opens the rewind menu, like /rewind.Shift+Tab cycles through the permission modes in turn, from the normal mode via automatically accepting edits to plan mode.! at the start runs a command directly in the shell, for example ! ls, and Claude responds to the result.
What to watch out for
- Lists on the web go stale quickly. An example from the docs: up to version 2.1.197,
/agents opened its own menu; today the command only reminds you to ask Claude for a subagent. When in doubt, what /help shows you is what counts.
- A missing command is often not an error. Some commands exist only on certain plans or systems.
Frequently asked questions
How do I see all Claude Code commands?
Type / into the input field. Claude Code then shows all commands available to you and filters the list as soon as you type more letters. /help shows the help with the available commands.
How do I continue an old conversation?
With /resume. Without an argument, it opens a picker with your earlier conversations; with a name or an ID you jump straight in. Give important sessions a name with /rename first, and you will find them more easily.
Can I create my own commands?
Yes. According to Anthropic, your own commands are skills: a file with instructions that you then call with /name. Claude Code already ships with some of these bundled skills itself, for example /doctor.
Why is a command from a tutorial missing for me?
Either it does not exist on your plan, system or version, or the tutorial is out of date. Claude Code hides commands that are not available to you. If you type one anyway, it reports an unknown command.
Where to go from here
Start your next session with /help, name it with /rename, and check once with /context what is already loaded before you give the first task. If you are still missing the basics of Claude Code, Claude Code: A Guide for Beginners is the step before this one. How others organise their sessions and AI employees is something we discuss in my German-speaking community Claude Practitioners.
