Claude Code slash commands

Show help for Claude Code

Displays a list of available commands, keyboard shortcuts, and pointers to documentation.

/help

Clear the conversation history

Wipes the current conversation transcript and starts a fresh session in the same project.

/clear

Local state like CLAUDE.md is preserved.

Compress conversation history to save tokens

Summarizes the prior conversation into a condensed form so you can keep working without hitting context limits.

/compact

Can be run with custom instructions for what to keep.

Resume a previous conversation

Lists prior sessions in the current project and lets you pick one to resume.

/resume

Switch which Claude model is active

Choose the Claude model used for this session, e.g. an Opus variant for harder work or a Haiku variant for speed.

/model

View or update Claude Code settings

Opens the configuration UI for things like default editor, theme, telemetry, and other user-level settings.

/config

For automated/programmatic changes prefer editing settings.json directly.

Initialize a new CLAUDE.md for the project

Scans the current codebase and produces a CLAUDE.md scaffold with project layout, commands, and conventions.

/init

Review a pull request

Performs a code review on a GitHub pull request, summarizing the diff and flagging issues.

/review 1234

Run a security review of pending changes

Audits the diff on the current branch for security issues like injection, secrets, and OWASP-style vulnerabilities.

/security-review

Show comments on a GitHub pull request

Fetches and displays the review comments on a GitHub PR so you can address them.

/pr_comments 1234

Show token usage and cost for the session

Displays cumulative token usage and estimated API cost for the current session.

/cost

Show Claude Code status

Prints the current version, model, working directory, and other diagnostic info about the session.

/status

Sign in to your Anthropic account

Authenticates the local Claude Code installation against your Anthropic account so API calls are billed correctly.

/login

Sign out of your Anthropic account

Clears local credentials. You'll need to /login again before sending API requests.

/logout

Diagnose Claude Code installation issues

Runs a series of environment and connectivity checks to help identify why Claude Code might not be working.

/doctor

Edit project memory (CLAUDE.md files)

Opens an editor on the project's CLAUDE.md and any nested CLAUDE.md files used as long-lived instructions.

/memory

Add a directory to the working set

Makes another directory available to Claude Code in this session, useful for monorepos.

/add-dir ../other-repo

Manage and switch between subagents

Lists registered subagents and lets you create or activate one. Subagents are specialized helpers like code review or planning.

/agents

Report a bug to Anthropic

Opens a bug-report flow that bundles relevant context like version and platform with your description.

/bug

Submit feedback about Claude Code

Sends free-form feedback to Anthropic about the Claude Code experience.

/feedback

Toggle Vim mode for input

Enables modal Vim-style editing in the Claude Code input box.

/vim

Manage hooks configuration

Hooks are shell commands the harness runs in response to events. This command opens the hooks editor.

/hooks

Hooks are persisted in settings.json.

Manage Model Context Protocol servers

Lists, adds, or removes MCP servers — external processes that provide tools and resources to Claude Code.

/mcp

View or edit tool permissions

Lists which tools Claude Code is allowed to invoke without prompting and lets you adjust the allowlist.

/permissions

View Claude Code release notes

Opens the changelog for the currently installed Claude Code version.

/release-notes

Upgrade Claude Code to the latest version

Checks for a newer Claude Code release and installs it in place.

Aliases: /update

/upgrade

Export the current conversation

Saves the current transcript to a file so you can share, archive, or post-process it.

/export

Set up terminal integration

Configures your terminal so Claude Code can detect window context and support multi-line input.

/terminal-setup

Connect Claude Code to your IDE

Pairs Claude Code with VS Code, JetBrains, or another supported IDE so they share state.

/ide

Show what's currently in your context window

Prints the files, tools, and memory entries currently loaded into the model's context.

/context

Configure the status line

Sets the command that produces the persistent status-line content at the bottom of Claude Code.

/statusline

View the current task list

Shows the in-session todo list Claude Code is tracking for the current task.

/todos

Show recent token usage and limits

Displays your recent API usage, rate-limit status, and remaining budget.

/usage

Migrate to the native Claude Code installer

Migrates a legacy npm-based Claude Code install to the new native installer.

/migrate-installer

Quit Claude Code

Ends the current Claude Code session and returns you to the shell.

Aliases: /quit

/exit

Run a prompt on a recurring interval

Schedules a slash command or prompt to be re-run on a cron-style cadence or self-paced loop.

/loop 5m /status

/loop /babysit-prs

From the superpowers plugin.

Schedule a recurring remote agent

Creates a cron-style schedule that runs a remote Claude Code agent on a routine.

/schedule 'daily at 9am' /status

Powered by remote-agent infrastructure.

Organize changes into logical commits

Groups staged and unstaged changes into a series of conventional commits, asking for confirmation before committing.

/commit

User-defined or plugin-provided; not built in.

Build, commit, and open a pull request

Bundles outstanding changes into logical commits, pushes the branch, and opens a GitHub PR with a summary.

/pr

Push the current branch to remote

Pushes the current Git branch, setting upstream tracking if needed.

/push

Create a new Git branch from a description

Formats your description into a clean conventional branch name and creates the branch.

/branch onboarding redesign

Bootstrap a new skill file

Scaffolds a new skill (Markdown + frontmatter) in the appropriate skills directory.

/init-skill my-helper

Part of the writing-skills workflow.

Multi-agent cloud code review

Runs a fleet of parallel reviewer agents against the current branch or a GitHub PR, then synthesizes their findings.

/ultrareview

/ultrareview 1234

Billed per run; user-triggered only.