Slash commands are control actions inside a session, and their full set is always wider than any static table. In an interactive session type a slash to see autocomplete; a given command's availability depends on the surface, provider, plan, plugins, MCP prompts and feature flags. User and project skills, as well as MCP prompts, add commands dynamically, so the runtime menu is the most accurate source, not memory. This reference gives a map of the most-used built-in commands by group; the full current list is on the official slash-commands page.
Removed commands are a separate topic, and restoring them by guesswork is not allowed. When a command disappears, it usually has a documented replacement named in the changelog and reference - and that is what you use. For example, the current branch removed /ultraplan, and /ultrareview became an alias of /code-review ultra; earlier separate flows like /vim and /pr-comments were removed. If a familiar command stopped working, the first step is not to find a workaround but to find the official replacement.
Commands are conveniently kept in mind by purpose, not as a list. There is a diagnostics-and-session group (/status, /doctor, /context, /resume, /rewind, /export), a model-and-context group (/model, /effort, /compact, /usage, /fast), a permissions-and-security group (/permissions, /sandbox, /login, /logout), an extensions group (/mcp, /agents, /config, /statusline, /keybindings) and a change-work group (/init, /memory, /review, /code-review, /run, /verify). Below is a map of the most-used of them.
The reference should be used with the same caveat as the whole book: a command's presence here is not a guarantee it exists on your surface and version, and its absence is not a sign it does not exist at all (skills and MCP add their own). Verify the actual menu with autocomplete, and a specific command's behavior with its own help. Fast-changing and cloud commands especially are worth checking against the official documentation before building a workflow on them.
| Group | Command | Purpose |
|---|---|---|
| Diagnostics/session | /status | Model, provider, directory, settings sources |
| /doctor | Check installation, settings, known issues (may fix) |
| /context | Context usage and memory files |
| /resume, /rewind | Resume a session; roll back to an earlier point |
| /export | Export the transcript outward |
| Model/context | /model, /effort | Choice of model and reasoning depth |
|---|
| /compact | Compress history now (say what to keep) |
| /usage, /fast | Limits and activity; fast mode |
| Permissions/security | /permissions, /sandbox | Resolved rules; sandbox state |
|---|
| /login, /logout | Login management |
| Extensions | /mcp, /agents | MCP status; create/edit subagents |
|---|
| /config, /statusline, /output-style | UI settings, status line, output style |
| Change work | /init, /memory | Starting CLAUDE.md; memory sources |
|---|
| /review, /code-review | Quick and deep review of a change |
| /run, /verify | Run and confirm a change by observation |