Claude Code is Anthropic's command-line tool for working with Claude, and slash commands are how you steer it: type / followed by a command to switch models, review code, manage your conversation, connect tools, or automate a task. This page is a plain-English map of the commands and core features worth knowing, with a dedicated guide behind each one.

How slash commands work

Every command starts with a forward slash and runs instantly, no full sentence required. Some open a picker or dialog, some take arguments (like /model opus), and some are skills: prompt-driven commands like /code-review and /loop that behave like built-in ones. The full command list lives in the official docs; the guides below explain the ones people ask about most, in language that assumes no background.

Type / in Claude Code to see the command menu.

The command reference

Pick the command or feature you're curious about. Each links to a short, practical guide.

Working sessions

• Claude Code Plan ModeWhat plan mode is and how it lets Claude research and propose an approach before it changes any code.
• /compact in Claude Code?How /compact summarizes a long conversation to free up context while keeping your project instructions intact.
• /clear in Claude Code?When and why to start a fresh conversation with /clear, and how it differs from /compact.
• Claude Code /resumeHow to pick a past conversation back up, or branch it to try a new direction without losing the original.
• Claude Code /rewind & Checkpoints ExplainedHow checkpoints let you roll the conversation and your code back to an earlier point when a change goes wrong.

Project & memory

• /init Do in Claude Code?How the /init command scans your project and writes a CLAUDE.md so Claude understands your codebase from the first session.
• Claude Code Memory & CLAUDE.md ExplainedHow CLAUDE.md and auto-memory give Claude persistent, project-specific instructions it loads at every session start.

Model & config

• Claude Code Output StylesWhat output styles are and how to switch or customize the way Claude formats and delivers its responses.
• Claude Code /modelHow to change which Claude model you're using, when to pick each tier, and how to set a default.
• Claude Code PermissionsHow allow, ask, and deny rules decide which actions Claude runs on its own and which need your approval.
• Claude Code /configHow to view and change Claude Code settings, from theme and model to editor mode and output style.

Tools & integrations

• Claude Code MCPWhat the Model Context Protocol is, and how to plug Claude Code into GitHub, Slack, databases, and the tools you already use.
• Claude Code SkillsWhat skills are, how the built-in ones work, and how to create your own so Claude does a task your way every time.
• Claude Code PluginsHow plugins package skills, hooks, MCP servers, and commands so teams can install a whole toolkit at once.
• Claude Code GitHub ActionsHow the GitHub integration lets Claude review pull requests, triage issues, and fix CI failures automatically.

Automation & scale

• /loop in Claude CodeRun a prompt or command on repeat, on an interval or self-paced, to poll, monitor, and automate recurring work.
• Claude Code HooksFire shell commands, HTTP calls, or prompts automatically at set points in a session, to enforce rules and integrate tools.
• Claude Code SubagentsHow subagents work in Claude Code, and how to use them to run focused, parallel tasks in isolated contexts.
• Claude Code Background Tasks & AgentsHow to detach long-running work to the background so you keep working while Claude handles it.
• Claude Code Headless ModeHow to run Claude Code non-interactively from the command line, a script, or a CI/CD pipeline.

Code quality

• Claude Code /code-reviewHow /code-review scans your changes for bugs and cleanups, and can post comments or apply fixes for you.
• Claude Code /security-review ExplainedHow to scan pending changes for security issues like injection, auth gaps, and data exposure before you ship.

Build on it

• Claude Agent SDKWhat the Agent SDK is and how it lets you build your own tool-using agents on the same engine as Claude Code.