Conversation & context

Start, compact, rewind and resume conversations; manage the context window

/clear [name]

Starts a new conversation with empty context while keeping project memory. The previous conversation is kept and can be reopened with /resume; pass a name to label it in the picker.

Best use: When you finish one feature and move to an unrelated task: /clear login-feature. Old context stops interfering and usage drops.

/compact [instructions]

Summarizes the conversation so far to free context space, continuing the same conversation. Optional instructions steer what the summary keeps.

Best use: After a long session: /compact keep the database schema and the list of open bugs.

/context [all]

Shows context usage as a colored grid with optimization suggestions and capacity warnings. all expands the per-item breakdown.

Best use: When Claude seems forgetful or usage climbs fast, run /context to see whether large files or tool output are filling the window.

/plan [description]

Enters plan mode, where Claude researches and proposes a plan before editing. Pass a description to start on that task immediately.

Best use: Before a big change: /plan fix the auth bug, review, then approve.

/resume [session]

Resumes a conversation by ID or name, or opens the session picker. Background sessions are marked bg.

Best use: You closed the laptop mid-task yesterday: /resume and pick yesterday's conversation.

/rewind

Rolls the conversation and/or code back to an earlier checkpoint, or summarizes from a selected message.

Best use: Claude broke several files: run /rewind, pick the message before the edits, and restore both code and conversation.

/autocompact [auto|<tokens>]

Sets how full the context window gets before automatic compaction, e.g. 500k; auto restores the model-tuned default. Saved to user settings and applied immediately.

Best use: Long task interrupted by frequent compaction: /autocompact 500k.

/branch [name]

Creates a branch of the conversation at this point and switches into it; the original is preserved and reachable via /resume. (A conversation branch, not a Git branch.)

Best use: Unsure between two approaches: /branch try-plan-b; if it fails, go back and continue plan A.

/btw [question]

Asks a side question about the current session without adding it to the conversation. Without a question, shows your most recent side question.

Best use: Mid-task you want to understand an error: /btw what does this 404 mean without derailing the main thread.

/copy [N]

Copies the last response to the clipboard; /copy 2 copies the second-to-last. With code blocks, a picker lets you copy one block; press w to write to a file instead (useful over SSH).

Best use: Claude drafted a snippet or command: /copy and paste it wherever you need it.

/export [filename]

Exports the conversation as plain text, directly to a file if you give a name, otherwise via a copy/save dialog.

Best use: Archive a valuable debugging session: /export debugging-notes.txt.

/goal [condition|clear]

Sets a goal: Claude keeps working across turns until the condition is met or the goal clears. No argument shows the current goal; clear (or stop, off, …) removes it.

Best use: /goal all tests pass, then let it iterate until the suite is green.

/memory

Edits CLAUDE.md memory files, toggles auto memory, and shows auto memory entries.

Best use: Want every answer in your language: open /memory and add a line such as “Always reply in Japanese.”

/recap

Generates a one-line summary of the current session on demand.

Best use: Back from a meeting and lost track: /recap.

/rename [name]

Renames the session and shows the name on the prompt bar; without a name, generates one from the conversation.

Best use: With many sessions open: /rename backtest so it stands out in /resume.

Model & reasoning

Switch models, reasoning effort and speed

/model [model]

Switches the model and saves it as the default for new sessions. No argument opens a picker; left/right adjusts effort on supported models; press s on a row to switch for this session only.

Best use: Use a faster model for small edits; switch with /model opus when a bug gets hard.

/advisor [model|off]

Enables or disables the advisor tool, which consults a second model at key moments. Accepts fable, opus, sonnet or a full model ID (fable requires access). No argument opens a picker.

Best use: Work on a faster model but get a stronger second opinion at key decisions: /advisor opus.

/effort [level|auto|status|ultracode [on|off]]

Sets the effort level: low to xhigh, max, or auto; status prints it. max is session-only. ultracode on|off toggles ultracode.

Best use: Hunting a flaky bug: /effort high; fixing a typo: /effort low.

/fast [on|off]

Toggles fast mode (same model, faster output). It can be toggled mid-response; the running turn finishes at its original speed. See the official fast mode docs for pricing.

Best use: Under time pressure: /fast on; otherwise /fast off.

/output-style [style]

Lists output styles or switches to one, e.g. /output-style concise.

Best use: Run /output-style with no argument to see what is available, then pick one (for example a more concise style).

Project & files

Set up project memory, add directories, review changes

/init

Initializes the project with a CLAUDE.md guide that every new session reads. Offers /import if it finds Codex or Gemini CLI configuration.

Best use: First thing in a new repo: /init, then refine it with /memory.

/add-dir <path>

Adds a working directory for file access this session without moving. Partial paths autocomplete with Tab. Most .claude/ config in the added directory isn't discovered, and most network paths can't be added.

Best use: Frontend and backend live side by side: /add-dir ../backend.

/cd <path>

Moves the session to a new working directory, keeping the conversation. Unlike /add-dir, /cd changes where the session lives.

Best use: Started in the wrong folder: /cd ~/projects/website instead of starting over.

/diff

Reviews changes in your working tree, including Claude's edits so far.

Best use: Before committing, /diff to see exactly what changed.

/import [codex|gemini|cursor] [--dry-run] [--yes]

Imports instruction files, MCP servers, commands, subagents and skills from OpenAI Codex, Gemini CLI or Cursor. --dry-run previews. Unavailable on some cloud providers; Cursor import requires v2.1.265+.

Best use: Moving from Cursor: /import cursor --dry-run first, then import for real.

/insights

Generates an HTML report of your recent sessions on this machine: projects, habits, where things go wrong, features to try. Not available in cloud sessions.

Best use: After a few weeks, run /insights to discover features you never tried.

/team-onboarding

Analyzes your last 30 days of sessions, commands and MCP usage and writes a Markdown guide a teammate can paste as their first message.

Best use: A new teammate joins: run /team-onboarding and send them the guide.

Review & verification

Find bugs and vulnerabilities, clean up code, run the app to verify

/code-review [low|medium|high|xhigh|max|ultra] [--fix] [--comment] [--max-findings n|all|default] [pr#|branch|path]

Reviews the current diff, or a PR, branch or path, for correctness bugs. --fix applies findings, --comment posts to the GitHub PR or GitLab MR, ultra runs a deep cloud review.

Best use: Before committing: /code-review high --fix. For a PR: /code-review high 1234.

/run

Launches and drives your project's app to see a change working.

Best use: /run to start the site and screenshot the new homepage.

/run-skill-generator

Writes a per-project skill that teaches /run and /verify how to build, launch and drive your app from a clean environment.

Best use: Startup needs several steps (database first, etc.): run this once, then /run just works.

/security-review

Analyzes your branch's diff against origin's default branch for vulnerabilities such as injection, auth issues and data exposure. Needs an origin remote.

Best use: Before shipping anything touching auth, payments or user data.

/simplify [target]

Four parallel agents review changed code for reuse, simplification, efficiency and abstraction level, then apply fixes. It does not look for correctness bugs; use /code-review for that.

Best use: The feature works and you want cleaner code: /simplify.

/ultrareview [PR or branch]

Deep multi-agent review in a cloud sandbox. The preferred form is /code-review ultra; /ultrareview is an alias. Includes 3 free runs on Pro, Max and Ultra, then requires usage credits.

Best use: Before a major release: /code-review ultra.

/verify

Builds, runs and observes your app to confirm a change works, rather than relying on tests or type checks. Runs only when you invoke it.

Best use: Changed a button's behavior: /verify so it actually clicks it.

Parallel & background work

Run several agents at once or detach work to the background

/agents

Prints a reminder to ask Claude to create or manage subagents, or to edit .claude/agents/ directly. On v2.1.197 and earlier, opened an interactive manager.

Best use: Just ask: “Create a subagent that only writes tests.”

/background [prompt]

Detaches the session to run as a background agent and frees the terminal. Optionally sends one more instruction first. Monitor with claude agents.

Best use: A 30-minute job: /bg write the results to report.md when done.

/batch <instruction>

Orchestrates large changes in parallel: researches the codebase, decomposes the work into 5–30 independent units and presents a plan; after approval, one background subagent per unit implements and tests it in its own worktree. Requires a git repository or a WorktreeCreate hook.

Best use: /batch migrate src/ from JavaScript to TypeScript.

/fork [prompt]

Copies the conversation into a new background session while you keep working here. With a prompt, the copy starts immediately; it is normally told to create its own worktree first.

Best use: Working on feature A and want a side fix in parallel: /fork fix the footer alignment.

/list-agents

Lists subagents, teammates and other sessions Claude can message, with the name to use for each. Only where cross-session messaging is enabled.

Best use: Coordinating several sessions: check what each is called.

/stop

Stops the background session you're attached to; transcript and worktree are kept. To detach without stopping, use /exit or press ←.

Best use: A background job went off course: /stop.

/subtask <task>

Spawns a forked subagent that inherits the conversation and reports its result back here. The difference from /fork: /subtask reports back here, while /fork starts a separate session.

Best use: /subtask document these API endpoints while you keep coding.

/tasks

Views and manages background work in this session, including finished subagents.

Best use: Check progress on background tests or subagents with /tasks.

/workflow-authoring

Loads the reference for writing dynamic workflow scripts. Claude normally loads it itself; run it before hand-editing a saved script. Only where dynamic workflows are enabled.

Best use: Before hand-editing a saved workflow script.

/workflows

Opens the workflow progress view to watch, pause, resume or save running and completed workflows.

Best use: Keep an eye on a long-running workflow.

Automation & scheduling

Repeat on an interval, run on a schedule, auto-fix pull requests

/loop [interval] [prompt]

Runs a prompt repeatedly while the session stays open. Omit the interval and Claude self-paces; omit the prompt and it runs the built-in maintenance prompt or your loop.md.

Best use: /loop 5m check if the deploy finished. For runs that continue when your computer is off, use /schedule.

/autofix-pr [prompt]

Spawns a cloud session that watches the current branch's PR and pushes fixes on CI failures or review comments. Requires the gh CLI and cloud sessions.

Best use: /autofix-pr only fix lint and type errors.

/deep-research <question>

A bundled workflow that fans out web searches, fetches and cross-checks sources, and writes a cited report. Runs in the background.

Best use: /deep-research compare pricing and quality of leading AI video tools.

/schedule [description]

Creates, updates, lists or runs cloud scheduled tasks (formerly routines) through a guided conversation. Not available with a Console API key.

Best use: /schedule every weekday at 9am, summarize new issues in the repo.

Artifacts, slides & design

Produce shareable pages, decks, designs and charts

/artifact-capabilities

Loads the reference for artifact runtime capabilities, such as calling connectors or offering downloads, and which your account has. Claude usually loads it itself.

Best use: Before building a page that many people fill in.

/artifact-diagramming

Loads diagramming guidance for artifacts: when a diagram helps, what to draw, and SVG that stays legible in light and dark themes.

Best use: When you need a clear architecture diagram.

/artifacts

Lists artifacts you own or that are shared with you, then attaches one, opens it, or copies its link. Where artifacts are available.

Best use: Find last week's dashboard and keep editing it.

/dataviz [request]

Chart design guidance: picks the chart form, assigns color by role, and validates the palette for colorblind safety and contrast with a bundled script.

Best use: /dataviz turn this monthly sales table into a trend chart.

/design [brief]

Drafts UI mockups, flows, landing pages or posters as artboards on one canvas (a Claude Design artifact); edit in a desktop browser and export PNG or PDF. Unavailable on Bedrock, Vertex and other third-party platforms.

Best use: /design a settings screen for a mobile banking app.

/design consent

Grants or withdraws Claude's access to Claude Design. When Claude asks you to run /design consent, type it yourself.

Best use: When prompted, type /design consent yourself.

/design-login

Authorizes design-system access for /design-sync with your claude.ai account, when /design-sync or /design consent asks you to.

Best use: Run it when /design-sync asks you to sign in.

/design-sync [hint]

Converts and uploads your repo's React design system to Claude Design so designs use your real components. First sync can take hours on large repos. It targets the standalone claude.ai/design site, which closes by December 14, 2026.

Best use: /design-sync Acme DS (mind the closing date above).

/slides [brief]

Makes a Claude Slides artifact from your brief. Requires a session with artifacts and an account where the Slides template is available.

Best use: /slides a quarterly review for leadership, 8 slides.

Plugins, skills & integrations

Extend Claude with plugins, skills, MCP servers, browser and IDE integrations

/mcp [reconnect (<server>|all)|enable|disable [<server>|all]]

Manages MCP server connections and OAuth. No argument opens the list; reconnect, enable or disable with a server name or all acts directly.

Best use: A connector failed or needs auth: /mcp reconnect all.

/chrome

Configures Claude in Chrome settings.

Best use: When first connecting the Chrome extension.

/claude-api [subcommand]

Loads Claude API and Managed Agents reference for your project's language; activates automatically when code imports anthropic. Subcommands include migrate, upgrade, api-version, cost-optimize, latency-optimize, build-eval, prompt-audit, managed-agents-onboard, managed-agents-secure, hillclimb, preserved-thinking-migration.

Best use: Cut API spend in your app: /claude-api cost-optimize.

/claude-in-chrome [task]

Has Claude carry out a task in your browser through Claude in Chrome, such as testing a page or reading console logs. Requires Chrome integration (e.g. claude --chrome).

Best use: /claude-in-chrome test the sign-up flow on my local site and report errors.

/ide

Manages IDE integrations and shows status.

Best use: Check the connection when edits don't show in your editor.

/install-github-app

Installs the Claude GitHub App for a repository, optionally setting up GitHub Actions workflows and secrets. github.com only.

Best use: To mention Claude in issues and PRs and have it work there.

/install-slack-app

Installs the Claude Slack app via a browser OAuth flow.

Best use: Your team wants Claude in Slack.

/plugin [subcommand]

Manages plugins. No argument opens the menu; list, install, enable, disable act directly. Run /reload-plugins if the install summary says so.

Best use: Browse with /plugin and install one that fits your work.

/plugin-authoring

Loads the reference for writing a mod. A skill from a built-in plugin, which you can turn off in /plugin.

Best use: When you want a small side panel that stays beside the chat.

/reload-plugins [--force]

Reloads active plugins without restarting and reports load errors. If the reload would change loaded MCP tools and invalidate the prompt cache, it warns and skips unless you pass --force.

Best use: Right after installing a plugin that asks for it.

/reload-skills

Re-scans skill and command directories so skills added or changed during the session are available without restarting.

Best use: You just wrote a custom skill and want it now.

/skill-doctor

Shows each skill's context cost and usage so you can find skills to turn off.

Best use: Run periodically to prune unused skills.

/skills

Lists skills with filtering. t sorts by token count, Space/Enter cycles a skill's visibility to Claude and the / menu, Esc saves and closes.

Best use: Too many skills: /skills, press t, hide the heavy unused ones.

Permissions & safety

Control what Claude may do and what needs your approval

/permissions

Manages allow, ask and deny rules for tools: view by scope, add or remove rules, manage working directories, and check the Recently blocked tab (calls auto mode refused) and classifier rules.

Best use: Tired of approving npm test every time: allow it here.

/auto-mode-setup

Drafts autoMode.environment entries from your project and recent sessions for you to review and save. Has version and plan requirements.

Best use: When auto mode keeps blocking routine actions.

/fewer-permission-prompts

Scans transcripts for common read-only Bash and MCP calls and adds a prioritized allowlist to project .claude/settings.json.

Best use: Run once when approval prompts keep interrupting you.

/hooks

Views hook configurations.

Best use: Check what's configured when a hook seems to break every edit.

/privacy-settings

Views and updates privacy settings. Pro, Max and Ultra only.

Best use: To check how your data is used.

/sandbox

Toggles sandbox mode, on supported platforms only.

Best use: Turn it on before running unfamiliar scripts.

/update-config [request]

Describe a settings change (allow a command, set an env var, add a hook) and Claude edits the matching settings.json. Use /config for theme and model.

Best use: /update-config run prettier after every file edit.

Settings & interface

Theme, colors, status line, keybindings, voice

/color [color|default]

Sets the prompt bar color for this session: red, blue, green, yellow, purple, orange, pink or cyan; default resets, no argument picks a random color.

Best use: Mark a production session with /color red.

/config [key=value ...]

Opens Settings for theme, model, output style and more, or sets key=value pairs directly, e.g. /config theme=dark. /config --help lists keys.

Best use: /config theme=dark.

/focus

Toggles a view showing only your last prompt, a one-line tool summary and the final response. Fullscreen rendering only.

Best use: Hide intermediate steps and see only the result.

/keybindings

Opens your keyboard shortcuts file.

Best use: Rebind actions to keys you're used to.

/scroll-speed

Adjusts mouse wheel scroll speed. Fullscreen rendering only; not in the JetBrains terminal.

Best use: Slow it down when one notch jumps too far.

/statusline

Configures the status line; describe what you want, or run without arguments to derive it from your shell prompt.

Best use: /statusline show the model and git branch.

/terminal-setup

Installs Shift+Enter for newlines in VS Code, Cursor, Devin Desktop, Alacritty or Zed; in Apple Terminal enables Option+Enter and disables the bell; in iTerm2 enables clipboard access for /copy.

Best use: Enter keeps sending half-written prompts: run this once.

/theme

Changes the color theme: auto, light and dark, colorblind-accessible, ANSI and custom themes.

Best use: Switch to a dark theme at night.

/tui [default|fullscreen]

Sets the terminal renderer and relaunches with the conversation intact. fullscreen is the flicker-free renderer. No argument shows the current renderer.

Best use: Screen flickers: /tui fullscreen.

/voice [hold|tap|off]

Toggles voice dictation or sets a mode (hold, tap). Requires a Claude.ai account.

Best use: Dictate instructions instead of typing.

Account & usage

Sign in, check limits and cost, upgrade

/usage

Shows session cost, spend for today, this week and this month, plan usage limits and activity stats, including what counts against plan limits on subscriptions. /stats opens the Stats tab. Each warning names its cause and a one-line fix; press f to apply the fixes.

Best use: Check /usage before deciding whether to switch to a cheaper model.

/login

Signs in to your Anthropic account.

Best use: New machine or expired session.

/logout

Signs out of your Anthropic account.

Best use: After using a shared computer.

/passes

Shares a free week of Claude Code with friends; only shown to eligible accounts.

Best use: Invite a friend to try it.

/rate-limit-options

Shows ways to keep working when a claude.ai limit blocks you: wait and auto-continue, add usage credits, or upgrade. Requires a claude.ai subscription.

Best use: Blocked overnight: choose auto-continue after reset.

/status

Opens Settings on the Status tab: version, model, account, connectivity. Works while Claude is responding.

Best use: Screenshot /status before asking for help.

/upgrade

Opens the upgrade page in your browser.

Best use: When you keep hitting limits.

/usage-credits

Configures or requests usage credits; Team and Enterprise members without billing access send a request to their admin. Previously /extra-usage.

Best use: Request credits from your admin when the team runs out.

Cloud, remote & other devices

Hand work between desktop, mobile, web and cloud

/desktop

Continues the session in the Claude Code Desktop app. Requires macOS or x64 Windows and a Claude subscription.

Best use: Move from the terminal to the desktop app.

/mobile

Shows a QR code to download the Claude mobile app.

Best use: To follow along from your phone.

/remote-control

Makes this session available for Remote Control from claude.ai. Requires a claude.ai subscription.

Best use: Leave the task running and steer it from your phone: /rc.

/remote-env

Chooses the default cloud environment for cloud sessions started from the CLI.

Best use: Pick a default when you have several environments.

/setup-bedrock

Configures Amazon Bedrock auth, region and model pins via a wizard. Hidden from the menu until CLAUDE_CODE_USE_BEDROCK=1 is set; type it in full.

Best use: Your company uses Claude through AWS.

/setup-vertex

Configures Google Cloud's Agent Platform via a wizard. Hidden until CLAUDE_CODE_USE_VERTEX=1 is set; type it in full.

Best use: Your company uses Claude through Google Cloud.

/teleport

Pulls a cloud session into this terminal, fetching its branch and conversation. Requires a claude.ai subscription.

Best use: Continue a web-started task locally.

/web-setup

Connects GitHub for cloud sessions using your local gh credentials.

Best use: Before your first cloud session.

Diagnostics & feedback

Start here when something is wrong

/doctor [prompt-audit [path]]

Runs a setup checkup for installation, settings, extensions and CLAUDE.md, proposing fixes Claude applies after you confirm. prompt-audit audits your instructions instead.

Best use: Something feels broken: start with /doctor.

/help

Shows help and available commands.

Best use: Forgot a command: /help, or type / plus a few letters.

/powerup

Interactive lessons with animated demos that teach Claude Code features.

Best use: The best structured starting point for newcomers.

/bug [report]

Reports a bug or shares your conversation; you choose how much history to include and confirm before anything is sent.

Best use: When Claude Code itself misbehaves.

/debug [description]

Enables debug logging from this point and troubleshoots by reading the session debug log. Optionally describe the issue.

Best use: /debug the MCP server keeps failing to connect.

/exit

Exits the CLI. In an attached background session it detaches and the session keeps running.

Best use: Done for the day.

/feedback [report]

Sends product feedback using the same dialog and consent step as /bug.

Best use: Request a feature.

/heapdump

Writes a heap snapshot and memory breakdown to the Desktop for diagnosing high memory. Share only the -diagnostics.json; the .heapsnapshot contains your conversation and credentials. Hidden from the menu.

Best use: Report a memory blow-up.

/release-notes

Views the changelog in a version picker; notes appear without entering Claude's context.

Best use: See what's new after an update.

/restart

Restarts Claude Code and resumes the session to apply an installed update. Requires Pro, Max or Ultra and a native macOS or Linux install (not WSL).

Best use: When the footer offers it after an update.

Other

Small extras

/radio

Opens Claude FM lo-fi radio in your browser, or prints the stream URL.

Best use: Background music while you work.

/stickers

Orders Claude Code stickers.

Best use: Decorate your laptop.

Removed

Still mentioned in older guides but no longer available

/pr-comments [PR] Removed

Removed in v2.1.91. Ask Claude directly to view pull request comments.

Best use: Ask in plain language.

/ultraplan <prompt> Removed

Removed. Use plan mode (/plan) instead.

Best use: Use /plan.

/vim Removed

Removed in v2.1.92. Use /config → Editor mode.

Best use: Use /config.