Claude Code slash commands: every command explained

Starts a new conversation with an empty context window. Your files, CLAUDE.md and memory stay. Pass a name to label the old conversation so you can find it in /resume. Also /reset and /new.

/clear

Replaces the conversation so far with a summary to free room in the context window. Add instructions to say what the summary must keep.

/compact keep the list of files changed and the test commands

Shows what is filling the context window as a coloured grid, with tips for what to trim. Add all to see every item.

/context

Sets how full the context window gets before Claude Code compacts on its own, as a size such as 500k, or auto for the default for your model.

/autocompact 400k

Asks a side question about the session without adding it to the conversation, so it costs no room in later turns.

/btw which file had the retry setting again?

Goes back to one of your earlier messages and restores the code, the conversation or both, or summarizes part of the conversation. Same as pressing Esc twice on an empty prompt. Also /checkpoint and /undo.

/rewind

Writes a one-line summary of the session so far.

/recap

Saves the conversation as plain text, to a file you name or through a dialog that copies or saves it.

/export auth-refactor.txt

Copies the last answer to the clipboard. Add a number for an older answer. When the answer has code blocks, you can pick one, or press w to write it to a file.

/copy 2

Reopens an earlier conversation by its name or id, or opens a list to pick from. Also /continue.

/resume oauth-migration

Names the current session so you can find it later. With no name, it makes one up from the conversation.

/rename oauth-migration

Copies the conversation at this point into a new branch and switches you to it, so you can try another approach. The original stays, and /resume takes you back.

/branch try-redis-cache

Copies the conversation into a new background session that works on its own, while you carry on here.

/fork write the migration for the new column

Sends the whole session to the background to keep working, and frees your terminal. Watch it with claude agents. Also /bg.

/background

Stops the background session you are attached to. Its conversation and any worktree are kept.

/stop

Moves this session to another folder and keeps the conversation.

/cd ../api

Lets Claude read and edit files in another folder as well, for this session. Most of that folder's .claude settings are not picked up.

/add-dir ../shared-lib

Brings a session that ran on the web into this terminal, with its branch and conversation. Needs a claude.ai plan. Also /tp.

/teleport

Lets you carry on this session from claude.ai or the Claude phone app, while it keeps running on your machine. Also /rc.

/remote-control

Moves the session into the Claude desktop app. macOS or Windows on x64, with a Claude plan. Also /app.

/desktop

Leaves Claude Code. In an attached background session it detaches and the session keeps running. Also /quit.

/exit

Switches to plan mode, where Claude reads and proposes a plan without editing your files. Add a task to start on it at once.

/plan add rate limiting to the public API

Switches the model and saves it as your default. With no name, opens a list; press s on a row to switch for this session only.

/model opus

Sets how much the model reasons before it answers, from low to max, or auto. status shows the current level.

/effort high

Turns fast mode on or off, which runs Opus faster at a higher price per token. On a Claude plan it is paid from usage credits, not your plan's limits.

/fast on

Turns on a second model that the main one can ask for advice at hard moments in a task, or turns it off.

/advisor opus

Sets a condition Claude keeps working towards across turns, such as all tests passing. A separate check runs after each turn. clear removes it.

/goal npm test passes with no failures

Lists the output styles or switches to one, which changes how Claude writes its answers.

/output-style explanatory

Shows the changes in your working tree, including what Claude has edited so far.

/diff

Reviews the current changes, or a pull request, branch or path you name, for bugs. Add a level from low to max, --fix to apply the findings, or ultra for a deeper review in the cloud. Also /review.

/code-review high --fix

Checks the changes on your branch against the main branch for security problems such as injection, broken access checks and leaked data.

/security-review

Has 4 reviewers look at the changed code for reuse, simplicity, efficiency and the right level of abstraction, then applies the fixes. It does not look for bugs.

/simplify

Runs a deep code review with several agents in a cloud sandbox. The same as /code-review ultra. A few free runs on Pro and Max, then paid.

/ultrareview 1234

Checks a change really works by building and running your app and watching the result, not only by running tests.

/verify

Starts your app and drives it so you can see a change working.

/run

Writes a skill for your project that teaches /run and /verify how to build and start your app from scratch.

/run-skill-generator

Starts a cloud session that watches your branch's pull request and pushes fixes when CI fails or a reviewer comments. Needs the gh tool.

/autofix-pr only fix lint and type errors

Splits a large change into 5 to 30 independent parts and, once you approve the plan, runs each in its own subagent and worktree.

/batch migrate src/ from JavaScript to TypeScript

Hands a side task to a subagent that starts with a copy of the conversation and reports back here when done.

/subtask find every caller of parseDate

Lists the background work in this session, including subagents and commands, finished or not. Also /bashes.

/tasks

Runs a prompt again and again while the session is open, at an interval you set or one Claude chooses. Also /proactive.

/loop 5m check whether the deploy finished

Creates, changes, lists or runs routines, prompts that run on a schedule in the cloud. Also /routines.

/schedule every weekday at 9 summarize new issues

Lists the subagents, teammates and other sessions Claude can send messages to. Also /peers.

/list-agents

Shows running and finished workflows, which spread a job across many subagents, to watch, pause, resume or save.

/workflows

Runs many web searches on a question, reads and cross-checks the sources, and writes a report with citations.

/deep-research how do other CLIs store login tokens?

Reads your project and writes a starting CLAUDE.md with its commands and conventions.

/init

Opens your CLAUDE.md files to edit, and turns auto memory on or off and shows what it saved.

/memory

Shows and edits your allow, ask and deny rules, and the folders Claude may work in. Also /allowed-tools.

/permissions

Looks through your past sessions for safe read-only commands you keep approving and adds them to the project's allow list.

/fewer-permission-prompts

Turns the sandbox on or off, which limits the files and websites shell commands can reach.

/sandbox

Drafts a description of your environment for auto mode's safety check from your project and recent sessions, for you to review and save.

/auto-mode-setup

Opens the settings screen for theme, model, output style and more, or sets a value directly. Also /settings.

/config theme=dark

Changes a settings file from a plain request, such as allowing a command or adding a hook.

/update-config allow npm run lint without asking

Brings instruction files, MCP servers, commands and skills over from OpenAI Codex, Gemini CLI or Cursor on your machine. --dry-run previews.

/import codex --dry-run

Checks your install, settings, extensions and CLAUDE.md for problems and offers fixes, which it applies after you confirm. Also /checkup.

/doctor

Sets up Shift+Enter for new lines in VS Code, Cursor, Alacritty or Zed, and fixes the equivalent settings in Apple Terminal and iTerm2.

/terminal-setup

Sets up the status line under the prompt. Describe what you want shown, or run it bare to copy your shell prompt.

/statusline show the branch and context used

Opens your keyboard shortcuts file to change keys.

/keybindings

Connects to your code editor and shows the connection.

/ide

Installs the Claude GitHub app on a repository and can set up the GitHub Actions workflow and its secret. github.com only.

/install-github-app

Installs the Claude app in Slack, through a sign-in in your browser.

/install-slack-app

Connects your GitHub account for cloud sessions, using the login of your local gh tool.

/web-setup

Picks the default cloud environment for cloud sessions you start from the terminal.

/remote-env

Changes the settings for Claude in Chrome, which lets Claude use your browser.

/chrome

Walks through signing in to Amazon Bedrock and choosing a region and models. Shows once CLAUDE_CODE_USE_BEDROCK=1 is set.

/setup-bedrock

Walks through signing in to Google Cloud and choosing a project, region and models. Shows once CLAUDE_CODE_USE_VERTEX=1 is set.

/setup-vertex

Lists your MCP servers and their status, signs in to the ones that need it, and reconnects, turns on or turns off a server.

/mcp reconnect github

Lists your skills, with what each costs in context. Press t to sort by size, and Space to hide a skill from Claude or from the menu.

/skills

Picks up skills you added or changed on disk during the session, without restarting.

/reload-skills

Shows what each skill costs in context and how often it is used, to find ones to turn off.

/skill-doctor

Reminds you how to create subagents - ask Claude, or edit the files in .claude/agents/ yourself.

/agents

Shows the hooks you have set up and where each comes from.

/hooks

Opens the plugin menu, or lists, installs, enables or disables a plugin directly.

/plugin install code-review@claude-plugins-official

Reloads your plugins so changes apply without a restart.

/reload-plugins

Loads the reference Claude follows to write a plugin for you.

/plugin-authoring

Loads the reference for writing workflow scripts that spread a job across many subagents.

/workflow-authoring

Loads reference material for building on the Claude API in your project's language, with tasks such as migrate, cost-optimize and build-eval.

/claude-api cost-optimize

Has Claude do a task in your browser, such as testing a page or reading its console, through Claude in Chrome.

/claude-in-chrome check the signup form on localhost:3000

Lists your artifacts, the web pages Claude publishes on claude.ai, to attach one to the session, open it or copy its link.

/artifacts

Loads the reference for what a published artifact page can do, such as offering a file to download.

/artifact-capabilities

Loads guidance for drawing clear diagrams in artifacts that work in light and dark themes.

/artifact-diagramming

Drafts screens, flows, landing pages or posters as artboards on a canvas you can edit in the browser and export.

/design a settings screen for a banking app

Uploads your repo's React design system to Claude Design, so its designs use your real components.

/design-sync

Signs in with your claude.ai account so /design-sync can upload your design system.

/design-login

Makes a slide deck from your brief as an artifact.

/slides a quarterly review of the platform team

Loads guidance for charts and dashboards - the right chart for the data, colours that work for colour-blind readers, and accessible labels.

/dataviz

Shows what the session has cost, how much of your plan's limits you have used, and activity stats. Also /cost and /stats.

/usage

Shows the version, model, account and connection, including which login or API key is in use. Works while Claude is answering.

/status

Signs in to your Anthropic account, or switches account.

/login

Signs out of your Anthropic account.

/logout

Shows your choices when a usage limit stops you - wait and carry on when it resets, add usage credits, or upgrade.

/rate-limit-options

Sets up usage credits to keep working past your plan's limits, or asks your admin for them.

/usage-credits

Opens the page to move to a higher plan.

/upgrade

Shows and changes your privacy settings. Pro and Max only.

/privacy-settings

Shares a free week of Claude Code with friends, if your account can.

/passes

Writes a report on your recent sessions on this machine - the projects, how you work, where things go wrong, and features to try.

/insights

Writes a guide for a teammate from how you have used Claude Code in the last 30 days, to paste as their first message.

/team-onboarding

Shows help and the commands you can use.

/help

Teaches Claude Code's features through short interactive lessons.

/powerup

Shows what changed in each version.

/release-notes

Turns on debug logging for the session and reads the log to find a problem. Describe the problem to focus it.

/debug MCP server keeps disconnecting

Reports a bug, with as much of the session as you choose to include, after a consent screen. Also /share.

/bug

Sends feedback about Claude Code, through the same dialog as /bug.

/feedback

Saves a memory snapshot to find out why Claude Code is using a lot of memory. Share only the diagnostics file, since the snapshot holds your conversation and credentials. Not shown in the menu; type it in full.

/heapdump

Changes the colour theme, including auto, colour-blind friendly themes and your own.

/theme

Sets the colour of the prompt bar for this session, which helps tell sessions apart.

/color green

Switches between the default and the full-screen display, keeping the conversation.

/tui fullscreen

In the full-screen display, shows only your last message, a one-line summary of the tools used and the final answer.

/focus

Changes how fast the mouse wheel scrolls in the full-screen display.

/scroll-speed

Turns voice dictation on or off, holding a key to talk or tapping to start and stop.

/voice hold

Shows a QR code to download the Claude phone app. Also /ios and /android.

/mobile

Opens Claude FM, a lo-fi radio stream, in your browser.

/radio

Orders Claude Code stickers.

/stickers