Claude Code slash commands: my cheat sheet, rechecked

My Notion list of Claude Code commands, checked line by line against the official docs for v2.1.294.

If you only learn six

/init creates a starter CLAUDE.md for the project. Once per project is enough.
/plan shows what Claude intends to touch before it touches anything.
/compact summarizes the conversation to free up context. It costs tokens itself, so don’t wait until you’re out of room.
/clear starts fresh. The old session isn’t lost; /resume brings it back.
/rewind rolls code and conversation back to an earlier checkpoint. It can even restore what was there before a /clear.
/usage shows what the session cost and where you stand against your plan limits.

Context and conversation

/context [all] draws your context usage as a colored grid, the quickest way to decide when to compact.
/btw [question] asks a side question without adding it to the conversation. With no question, it shows your earlier ones.
/recap gives a one-line summary of the session.
/export [file] saves the conversation as plain text.
/copy [N] copies the last response, or the Nth-latest, with a picker for individual code blocks.
/rename [name] names the session. Leave the name out and it makes one up.
/autocompact [auto|tokens] sets how full the context gets before it compacts on its own, e.g. 500k. Remembered per model.
/focus hides the noise: your last prompt, a one-line tool summary and the final answer.
/output-style [style] lists or switches output styles.

Models and effort

/model [model] switches model and saves it as the default. Press s in the picker to switch for this session only.
/effort [level|auto|status] sets how hard it thinks: low, medium, high, xhigh, max or auto. max doesn’t persist past the session.
/effort ultracode [on|off] adds automatic workflow orchestration on top of deep reasoning.
/advisor [model|off] lets a second model weigh in at key moments.
/fast [on|off] toggles fast mode.

Working on its own

/goal [condition|clear] keeps Claude going until a condition is true, like “tests pass”. With no argument it shows the current goal.
/loop [interval] [prompt] repeats a prompt while the session is open, e.g. /loop 5m check if the deploy finished. Skip the interval and Claude paces itself.
/workflows opens the progress view for running workflows, where you can pause, resume or save them.

Parallel agents

/background [prompt] detaches the session as a background agent (/bg). /tasks lists what’s running.
/branch [name] forks the conversation so you can try another direction. /resume takes you back.
/fork [prompt] copies the conversation into a new background session and keeps you where you are.
/subtask <task> hands a side task to a background subagent that inherits the whole conversation and reports back into this one.
/batch <instruction> breaks a big change into 5–30 independent pieces and runs a subagent per piece in its own worktree. You approve the plan first.
/list-agents shows which subagents, teammates and sessions you can message, and the name to use for each (/peers).
/deep-research <question> runs parallel web searches, cross-checks the sources and returns a cited report.
/schedule sets up cloud routines (/routines), and /autofix-pr watches your branch’s PR and fixes CI failures and review comments.

Reviewing and shipping

/diff is an interactive viewer for uncommitted changes and per-turn diffs.
/code-review [level] [--fix] [--comment] [--max-findings n|all|default] [target] looks for bugs and cleanups. --fix applies what it finds, --comment posts to the PR, and ultra runs the deep cloud review. Your --max-findings choice sticks until you set it back to default. /review is an alias.
/simplify [target] cleans up changed code for reuse, simplicity and efficiency. It doesn’t hunt bugs.
/security-review is a read-only security pass over the pending changes.
/ultrareview [PR] is the same as /code-review ultra. Pro and Max get three free runs.

Sessions

/resume [session] reopens an older conversation by ID, name or picker.
/stop stops a background session you’re attached to. The transcript and worktree stay.
/exit leaves the CLI. Inside a background session it only detaches.

Settings, permissions and memory

/config is the settings screen (theme, model, editor mode). It also takes key=value, e.g. /config theme=dark.
/permissions manages allow, ask and deny rules.
/fewer-permission-prompts scans your transcripts and builds an allowlist from the read-only calls you keep approving.
/update-config [request] takes a plain description, like “allow npm test”, and edits the right settings.json.
/memory edits your CLAUDE.md files and auto-memory. /hooks shows your hooks.
/add-dir <path> lets Claude read another directory without moving the session. /cd <path> actually moves it.
/import [codex|gemini|cursor] brings over instruction files, MCP servers, commands, subagents and skills from those tools. Add --dry-run to preview.
/team-onboarding writes a getting-started guide for a teammate from your last 30 days of usage.
/auto-mode-setup drafts autoMode.environment entries from your project and recent sessions, and saves them after you review.
/sandbox toggles sandbox mode where the platform supports it.

Skills, plugins and MCP

/skills lists skills. t sorts by token cost.
/skill-doctor shows what each skill costs in context and how often it’s used, so you can switch off the dead weight.
/reload-skills and /reload-plugins pick up changes without a restart. /plugin manages plugins.
/run starts your app and drives it to see a change actually work, and /verify goes further and checks the behavior, not just the tests. /run-skill-generator writes the project skill they rely on.
/mcp manages MCP connections and OAuth. /ide and /chrome handle the IDE and Claude in Chrome.
/plugin-authoring writes a mod, a plugin that changes Claude Code from the inside. Mods have their own post.

Custom commands are written as skills (.claude/skills/). Old files in .claude/commands/ still work, and if a skill and a command share a name, the skill wins.

Account and limits

/login, /logout.
/rate-limit-options is what you open when a usage limit blocks you: wait for the reset, add credits, or upgrade.
/usage-credits sets up credits (or requests them from your admin), and /upgrade opens the plan page.
/privacy-settings is for your privacy options on Pro and Max.

Everything else

Between devices: /teleport pulls a web session into the terminal, /desktop continues in the desktop app, /mobile shows a QR code, and /remote-control (/rc) makes the session reachable from claude.ai.
Terminal and looks: /terminal-setup (Shift+Enter), /theme, /statusline, /keybindings, /color, /voice for dictation, and /tui fullscreen for the flicker-free renderer.
Setup and diagnosis: /doctor checks your setup and proposes fixes you confirm first (/doctor prompt-audit checks your instructions instead). /debug turns on debug logging. /heapdump writes a memory snapshot, and only the -diagnostics.json file is safe to share. The snapshot holds your whole conversation.
Cloud and apps: /setup-bedrock, /setup-vertex, /install-github-app, /install-slack-app, /web-setup, /remote-env.
Bundled skills: /claude-api, /dataviz, /design, /design-sync, /slides, /workflow-authoring.
Discovery: /insights builds a report on how you use Claude Code, /powerup has short interactive lessons, /release-notes and /status do what they say, and /artifacts lists your artifacts.
Odds and ends: /bug and /feedback, /radio, /stickers, /passes.

If a command misbehaves, /help shows what your version actually has, and the official reference is the source of truth. Commands change almost every release.