Command Overview
A purpose-oriented catalog for choosing Claude Code commands quickly.
Key takeaways
- A purpose-oriented catalog of high-signal Claude Code commands, marked with the v2.1-era release that introduced or changed each one (checked 2026-06-25 against v2.1.190 latest and v2.1.179 stable).
- A command is only recognized at the start of a message; trailing text becomes its arguments, and typing
/lists everything available to you. - Selection rules disambiguate overlapping commands:
/usageover/costand/stats,/code-review ultraover/ultrareview,/cdversus/add-dir, and/branchversus/fork. - Use
/model fable(orbest) only for the hardest autonomous work, andclaude --safe-modeto check whether repo context triggered a fallback. /schedulehandles unattended cloud automation while/loophandles local session loops;/backgroundand/workflowscover different long-running cases.
Scope
The official Commands page lists every command in Claude
Code, and notes that availability varies by platform, plan, and environment. This catalog focuses on
high-signal commands for real repository work and marks recent v2.1-era changes where they affect
operating practice. Where a row repeats a fact straight from the reference, it uses the wording the
docs use today. Checked on 2026-06-25; the latest Claude Code release covered here is v2.1.190,
with npm stable at v2.1.179.
A command is only recognized at the start of a message; any text after the command name is passed to
it as arguments. Type / to list everything available to you.
Catalog
| Command | Status | Use it for |
|---|---|---|
claude mcp login/logout | v2.1.187 CLI | Refresh or clear MCP OAuth auth from scripts or setup runbooks. |
CLAUDE_CODE_MCP_TOOL_IDLE_TIMEOUT | v2.1.185 env | Tune idle timeout for long-running MCP tool calls. |
sandbox.credentials | v2.1.186 setting | Tighten credential boundary handling inside sandboxed execution. |
/advisor [model|off] | v2.1.170 Fable update | Consult a second model; use fable when your organization has Fable 5 access, or turn it off with off. |
/cd <path> | v2.1.169 | Switch the session working directory while preserving prompt cache. The new directory's CLAUDE.md is appended as a message, session storage relocates there, and unknown directories trigger a trust prompt. |
/model | v2.1.170 model baseline | Switch the model and save it as the default; supports fable, best, opus, sonnet, and haiku aliases. Use left/right to adjust effort, or s for session-only. |
/workflows | v2.1.154 | Watch, pause, resume, or save dynamic workflow runs. |
/fast | v2.1.154 update | Toggle fast mode on or off (/fast [on|off]) when latency matters more than cost. |
claude --bg --exec '<command>' | v2.1.154 CLI | Run a shell command as an Agent View background job without invoking a model. |
/code-review | v2.1.147+ | Review the diff for correctness bugs and cleanups. Effort: low|medium|high|xhigh|max|ultra; add --fix or --comment. /code-review ultra runs the deep cloud review. |
/run | v2.1.145+ | Launch and drive the app to see the change working (skill; requires v2.1.145+). |
/verify | v2.1.145+ | Confirm a change by building, running, and observing the app (skill; requires v2.1.145+). |
/run-skill-generator | v2.1.145+ | Write a per-project skill that teaches /run and /verify how to build and drive the app. |
/goal | v2.1.139 | Set a completion condition so Claude keeps working across turns; clear/stop/off ends it early. |
/background, /bg | v2.1.136+ | Detach the current session to run as a background agent and free the terminal. |
/usage-credits | v2.1.134 | Configure usage credits to keep working when you hit a limit. Previously named /extra-usage. |
/reload-skills | v2.1.152 | Re-scan skill and command directories so on-disk changes apply without restarting. |
claude project purge [path] | v2.1.126 CLI | Delete local Claude Code project state after previewing with --dry-run. |
/skills | v2.1.121 UX update | List available skills, sort by token count (t), and hide a skill from Claude or the / menu. |
/theme | v2.1.118 update | Select built-in, auto, daltonized, ANSI, or custom UI themes. |
/usage (/cost, /stats) | v2.1.118 consolidation | The standard cost and quota command. /cost and /stats are aliases; /stats opens on the Stats tab. |
/fewer-permission-prompts | v2.1.111 skill | Scan transcripts and add a prioritized allowlist for repeated read-only Bash and MCP calls to .claude/settings.json. |
/focus | v2.1.110 | Toggle the focus view down to the latest prompt, a one-line tool summary, and the final response (fullscreen only). |
/tui | v2.1.110 | Switch between the default and fullscreen renderers and relaunch into it. |
/recap | v2.1.108 | Generate a one-line summary of the current session on demand. |
/team-onboarding | v2.1.101 | Generate a team onboarding guide from the past 30 days of Claude Code usage. |
/powerup | v2.1.90 | Discover Claude Code features through quick interactive lessons. |
/branch | v2.1.77 rename | Branch the current conversation; /fork is an alias unless CLAUDE_CODE_FORK_SUBAGENT is set. |
/ide | v2.1.73 | Manage IDE integrations and show status. |
/effort | v2.1.72 | Tune effort: low, medium, high, xhigh, max, or ultracode (where available); auto resets to default. |
/loop | v2.1.71 | Re-run a prompt repeatedly while the session stays open; omit the interval to let Claude self-pace. |
/color | v2.1.70 | Set a session prompt-bar color, default to reset, or no argument to pick a random color. |
/rename | v2.1.70 | Name the current session for later discovery; no argument auto-generates one. |
/claude-api | v2.1.69 skill | Load Claude API, SDK, and Managed Agents guidance; migrate and managed-agents-onboard subcommands. |
/reload-plugins | v2.1.69 | Reload active plugin components without restarting and report load errors. |
/batch | v2.1.63 skill | Decompose large codebase work into 5-30 independent units, each run in its own git worktree. |
/simplify | v2.1.63, v2.1.154 update | Cleanup-only review (reuse, simplification, efficiency, abstraction) that applies fixes; from v2.1.154 it does not hunt for correctness bugs. |
/copy | v2.1.59 | Copy the last response, the Nth-latest (/copy 2), or a selected response/code block. |
/debug | v2.1.30 skill | Enable debug logging from this point and investigate session issues. |
/keybindings | v2.1.18 | Open or create the keybindings configuration file. |
/doctor | v2.1.3 | Diagnose installation and settings; press f to have Claude fix reported issues. |
/plan | v2.1.0 | Enter plan mode directly; pass a description to start on a task immediately. |
/teleport | v2.1.0 | Pull a Claude Code on the web session into the terminal (also /tp). |
/remote-env | v2.1.0 | Configure the default remote environment for web sessions started with --remote. |
/plugin | v2.0.12 | Install, enable, validate, and manage plugins. |
/permissions | v2.0-era | Manage tool allow, ask, and deny rules (alias /allowed-tools). |
/mcp | v1.0.18 | Manage MCP server connections and OAuth status. |
/add-dir | v1.0.42 | Add a working directory for file access during the session. |
/context | v1.0.86 | Visualize context usage and show optimization suggestions; all expands the breakdown. |
/terminal-setup | v1.0.110 | Configure terminal keybindings such as Shift+Enter for supported editors. |
/config (/settings) | Current docs | Open the Settings interface for theme, model, output style, and editor mode. /settings is an alias. |
/init | Current docs | Initialize a CLAUDE.md guide; it suggests improvements rather than overwriting an existing file. CLAUDE_CODE_NEW_INIT=1 enables an interactive flow covering skills, hooks, and personal memory. |
/autofix-pr | Current docs | Spawn a web session that watches the branch's PR and pushes fixes when CI fails or reviewers comment. |
/schedule | Current docs | Create or manage Claude Code on the web routines (alias /routines). |
/desktop | Current docs | Continue the current session in Claude Code Desktop on macOS or Windows (alias /app). |
/web-setup | Current docs | Connect GitHub for Claude Code on the web using local gh credentials. |
/ultrareview | Legacy alias | Deep multi-agent cloud review. Prefer /code-review ultra; /ultrareview remains as an alias. |
/pr-comments | Removed v2.1.91 | Gone. Ask Claude directly to view pull request comments instead. |
/vim | Removed v2.1.92 | Gone. Toggle Vim/Normal editing via /config → Editor mode. |
/extra-usage | Renamed | Old name for usage-credit flows; use /usage-credits. May still appear in older docs or transcripts. |
Selection Rules
Use /usage as the operating default for cost and quota checks. /cost and /stats are aliases of
/usage; reach for them only as shortcuts, with /stats opening directly on the Stats tab.
Use /model fable only for the hardest or longest-running autonomous work. best selects Fable when
available and otherwise the latest Opus. If Fable falls back before an unusual prompt, start
claude --safe-mode to check whether repo context or custom configuration is triggering classifiers.
Use /add-dir when you only need file access to another directory. Use /cd when the session itself
should move, including future --resume and --continue lookup from the new directory.
Use /usage-credits to continue past a limit. /extra-usage is the previous name for the same flow
and may still appear in older docs or transcripts.
Use /config to open Settings; /settings is an alias for the same interface. Vim editing now lives
here under Editor mode rather than behind the removed /vim command.
Use /branch in team docs instead of /fork. /fork is an alias of /branch in normal mode, but
when CLAUDE_CODE_FORK_SUBAGENT is set, /fork instead spawns a forked subagent and is no longer an
alias.
Prefer /code-review ultra for deep cloud review. /ultrareview still works as an alias, but the
combined /code-review entry point exposes the full effort range (low through max, plus ultra)
and the --fix and --comment flags in one place.
Use /code-review to find correctness bugs and /simplify for cleanup-only passes. As of v2.1.154
/simplify no longer looks for bugs; on earlier versions it behaved like /code-review --fix.
For setup, /init generates a starter CLAUDE.md and, on an existing project, suggests improvements
rather than overwriting it. Set CLAUDE_CODE_NEW_INIT=1 for the interactive flow that also walks
through skills, hooks, and personal memory files.
Use skills for reusable workflows, not only .claude/commands files. Existing command files keep
working, but skills support invocation controls, supporting files, and safer permission scoping.
Use /schedule for unattended cloud automation and /loop for local session loops. They solve
different scheduling problems.
Use /background and Agent View for local long-running sessions you may attach to later. Use
/workflows when Claude is orchestrating a generated workflow across many background agents.