Skip to main content
Slash commands are short directives you type directly in the Claude Code prompt, prefixed with /. They control session state, configuration, integrations, and tooling — without consuming any AI tokens. Claude never sees them; they are intercepted and handled locally before the message is sent to the model. 
# Type a slash command at any Claude Code prompt
> /clear
> /compact focus on the authentication logic
> /model claude-opus-4-5
Commands are available in the interactive REPL (claude). Most are not available in non-interactive / pipe mode unless the entry in the table below notes supportsNonInteractive: true.

​
All commands by category

CommandAliasesDescription
/clearreset, newClear conversation history and free up context
/compact [instructions]—Summarise conversation history and continue with a compact context
/resume [id or search]continueList and resume a previous conversation session
/context—Visualise context window usage as a coloured grid
/cost—Show total API cost and duration for the current session
/usage—Show plan usage limits (claude.ai subscribers)
/export—Export current conversation to a file
/rewind—Rewind to an earlier point in the conversation
/exitquitExit the REPL
CommandAliasesDescription
/configsettingsOpen the interactive configuration panel
/model [model]—Switch the active AI model
/memory—Edit CLAUDE.md project memory files
/plan [open|description]—Enable plan mode or view the current session plan
/theme—Change the UI colour theme
/vim—Toggle between Vim and Normal editing modes
/effort [level]—Set reasoning effort level (low, medium, high, max, auto)
/status—Show version, model, account, API connectivity, and tool status
/doctor—Diagnose and verify your Claude Code installation
/permissionsallowed-toolsManage allow/deny tool permission rules
/hooks—View hook configurations for tool events
/keybindings—View and manage keyboard shortcuts
/sandbox—Toggle or configure sandbox restrictions
/init—Create or improve a CLAUDE.md for the current repository
CommandAliasesDescription
/mcp [enable|disable [server]]—Manage MCP server connections
/login—Sign in to your Anthropic account
/logout—Sign out from your Anthropic account
/install—Install or reinstall the Claude Code native binary
/install-github-app—Set up Claude GitHub Actions for a repository
/install-slack-app—Install the Claude Slack app
/desktopappContinue the current session in Claude Desktop
/ide—Connect to an IDE integration
/feedback [report]bugSubmit feedback or report a bug
/upgrade—Upgrade to Max plan for higher rate limits (claude.ai)
CommandAliasesDescription
/agents—Manage agent configurations
/review [PR number]—Review an open pull request
/tasks—View and manage background tasks
/add-dir [path]—Add an additional working directory
/branch—Manage git branches
/diff—Show current git diff
/skills—View and manage loaded skills
/plugin—Manage Claude Code plugins
CommandAliasesDescription
/help—Show help and available commands
/version—Print the current Claude Code version
/status—Show full runtime status summary
/doctor—Run installation diagnostics
/release-notes—View release notes for the current version
/insights—Generate a usage analytics report for your sessions
/stats—Show session statistics

​
How slash commands work

1

Input interception

When you press Enter after typing /command, Claude Code intercepts the input before it reaches the AI model. The command name is matched against the built-in command registry.
2

Argument parsing

Everything after the command name is passed as a raw argument string to the command handler. Commands that accept arguments document them with an argumentHint, visible in autocomplete.
3

Local execution

The command runs locally — modifying session state, opening a UI panel, or performing a system operation. No API call is made unless the command itself triggers one (e.g. /compact).
4

Return to prompt

After the command completes, the REPL returns to the input prompt, ready for the next message or command.

​
Custom slash commands (skills)

In addition to built-in commands, Claude Code loads skills from:
  • .claude/commands/ in the current project directory
  • ~/.claude/commands/ in your home directory
Skills are Markdown or MDX files whose filename becomes the command name. Place them in .claude/commands/ to make them available as slash commands. 
.claude/
└── commands/
    ├── deploy.md          # available as /deploy
    └── generate-tests.md  # available as /generate-tests

Build docs developers (and LLMs) love

Get started for freeTalk to us