Slash commands, surface by surface
Slash commands, mapped
by coding agent and surface.
GitHub Copilot, Claude Code, and OpenAI Codex all
answer to / — and each surface exposes a different command set. Compare desktop
apps, terminals, editor integrations, and documented coding-agent web experiences, without mixing
in the general Claude or ChatGPT composer menus.
Pick a surface below, or jump to Compare to see who has what.
No commands match that search.
Where each command lives
Same name, different product — and sometimes different behavior. A filled cell links to an indexed command; read its badges for inherited, experimental, or unavailable behavior on that surface. A faint middle dot means the command is not listed. In columns marked documented subset or availability varies, a question mark means coverage or runtime support is not established, rather than proving that the command is unavailable.
About this reference
Why split by product and surface?
Because typing / gives you a different menu depending on where you are — and
“where you are” has two parts: which agent, and which of its surfaces.
Within GitHub Copilot alone the sets diverge sharply. The desktop app is built on
top of the GitHub Copilot CLI, but that shared engine does not imply identical menus. Both document
names such as /chronicle, /compact, and /model; the app adds
workflows like /orchestrate, /create-canvas, and /inbox instead
of exposing every terminal command. GitHub Copilot Chat in VS Code also has editor-specific actions:
editor actions (/fix, /tests, /explain) plus customization
authoring (/create-skill, /create-agent). The JetBrains, Visual Studio
and Xcode entries cover their published subsets, not every CLI-backed or preview agent integration.
Across products the overlap is real but shallow. Several names — /init,
/compact, /model, /review — show up in more than one
agent, usually meaning roughly the same thing. Others collide outright: the same token does
different work depending on whose prompt box you are in. The
Compare view exists to make both cases visible.
What counts as a surface?
Slash Command Atlas catalogs commands in coding-agent interfaces using first-party documentation,
source code, and explicitly qualified observations. A browser
counts only when the vendor documents commands for the coding product itself. That includes
GitHub Copilot Chat on GitHub.com and Claude Code on the web at
claude.ai/code. It excludes ordinary Claude chat: its / menu contains general
composer options and account skills, not Claude Code commands. No Codex web slash-command
inventory is included here. ChatGPT web has its own contextual composer menu, and the
desktop and CLI reference does not apply there. This is a boundary on the Atlas's evidence, not a
claim that Codex cloud workflows or ChatGPT web commands do not exist.
Editor integrations are separate surfaces only when their composer exposes a distinct documented set. The Codex IDE extension qualifies with an explicit slash-command table. Claude Code in VS Code is included as a documented subset because Anthropic says the extension exposes only part of the CLI set; Claude Code’s JetBrains plugin stays under CLI because it launches and connects the terminal interface.
A note on naming
This reference says GitHub Copilot, never just “Copilot”. Microsoft
ships several unrelated products under that word — Microsoft 365 Copilot, Copilot in Windows,
Security Copilot and others — and none of them share this command set. Claude Code
means Anthropic’s coding agent, including its Code tab and claude.ai/code, not an
ordinary Claude chat. OpenAI Codex means the current coding agent, not the general
ChatGPT web composer or the deprecated Codex model API of the same name.
Reading the badges
- Requires The command is greyed out or hidden until this condition is met — usually an active session, a repository, or an open pull request.
- Built-in skill The command runs an agent skill rather than a hard-coded action, so its behavior can evolve without a client update.
- Bundled workflow The command orchestrates a packaged, multi-agent workflow rather than one fixed UI action.
- Dynamic command The token is a pattern populated by your own prompts, skills, plugins, or connected servers.
- Hidden from menu The command works when typed in full but is intentionally omitted from the normal picker.
- Preview Publicly available but explicitly labelled as subject to change.
- Experimental Off by default or gated behind an experimental flag.
- Inherited built-in The vendor says this surface inherits built-ins from another surface but does not publish per-command behavior for every inherited entry.
- Not available here The vendor documents the token but says this surface refuses it or requires a different interface.
- Alias A second name for the same command.
Sources
Command names and behavior are checked against first-party documentation and source code. Source-backed, inherited, and previously observed entries carry qualifications where the published surface reference is incomplete; a documentation review is not a fresh runtime check of every command. The longer explanations, use cases, and example prompts are editorial — written to make the reference usable, not copied from the docs.
Staying current
These products change their command sets frequently, and availability also depends on your plan,
your organization’s policies, platform, and client version. This site is a snapshot. The
authoritative answer for your setup is always the same: type / in the prompt
box and read the picker.