Codex CLI slash commands · AI coding agents

Slash commands are typed into the TUI composer. Type / to open the command popup; pressing Tab while Codex is working queues a command for the next turn.

For terminal-level commands like codex exec, see Codex CLI terminal commands. For composer shortcuts (@, !, Esc Esc, etc.), see the notes at the bottom of this page.

Full command reference

Command	What it does	When to use it
`/init`	Generate an `AGENTS.md` scaffold in the current directory	First session in a new repo
`/status`	Show session config and token usage (model, approval policy, writable roots, context left)	Checking what Codex can see/do right now
`/permissions`	Set what Codex can do without asking (approval presets, custom permission profiles)	Changing autonomy mid-session
`/model`	Choose model and reasoning effort	Harder task needs more reasoning; cheaper model for routine work
`/review`	Review the working tree (uses session model unless `review_model` is set)	Before committing local changes
`/new`	Start a new chat in the same CLI session (`/new <name>` to name it)	Switching to an unrelated task
`/clear`	Clear the terminal and start a fresh chat	Full reset (unlike Ctrl+L, which only clears the view)
`/compact`	Summarize the chat to free context	Long session approaching the context limit
`/diff`	Show the Git diff, including untracked files	Reviewing what Codex changed
`/mention`	Attach a file to the chat	Pointing Codex at a specific file
`/mcp`	List configured MCP tools; `/mcp verbose` for server diagnostics	Verifying MCP servers are connected
`/logout`	Sign out and clear local credentials	Switching accounts
`/quit`, `/exit`	Exit the CLI	Done for the day
`/feedback`	Send logs/diagnostics to the maintainers	Reporting a bug
`/ps`	Show background terminals and recent output (requires `unified_exec`)	Checking on long-running commands
`/stop`	Stop all background terminals (`/clean` is an alias)	Killing stuck background processes
`/plan`	Switch to plan mode, optionally with an inline prompt	Want a plan before any edits
`/goal`	Set/edit/pause/resume/view/clear a persistent task goal (max 4,000 chars)	Multi-step objectives with automatic continuation
`/resume`	Resume a saved chat from the picker	Returning to earlier work
`/fork`	Fork the current chat into a new chat	Trying an alternative approach without losing the transcript
`/agent`, `/subagents`	Switch the active agent thread	Inspecting or continuing spawned subagent work
`/skills`	Browse and use skills	Running a packaged workflow
`/apps`	Browse apps (connectors), insert as `$app-slug`	Pulling in a connected app
`/plugins`	Browse installed/discoverable plugins (Space toggles enabled)	Managing plugins
`/hooks`	View/manage lifecycle hooks	Debugging or editing hooks
`/ide`	Include IDE context (open files, selection)	Using what you have open in the editor
`/keymap`	Remap TUI keyboard shortcuts (persists to `tui.keymap`)	Customizing key bindings
`/vim`	Toggle composer Vim mode	Prefer modal editing in the composer
`/personality`	Set communication style (`friendly`, `pragmatic`, `none`)	Adjusting response tone
`/fast`	Toggle the Fast service tier	Only shown if the model catalog advertises one
`/copy`	Copy the latest completed output (also Ctrl+O)	Pasting a result elsewhere
`/raw`	Toggle raw scrollback mode (also Alt+R; `tui.raw_output_mode`)	Selecting/copying unformatted output
`/rename`	Rename the current chat	Organizing saved sessions
`/archive`	Archive the current session and exit	Hiding finished work
`/delete`	Permanently delete the current session and exit	Removing a session entirely
`/side`, `/btw`	Ephemeral side chat	Quick question without polluting the main chat
`/app`	Continue the session in the ChatGPT desktop app (macOS/Windows)	Moving to the desktop UI
`/experimental`	Toggle experimental features	Trying preview functionality
`/approve`	Approve one retry of a recent auto-review denial	Overriding a single auto-review decision
`/memories`	Configure memory use/generation	Managing what Codex remembers
`/import`	Import Claude Code or Cursor setup/projects/chats (up to 50 chats from the last 30 days)	Migrating from another tool
`/usage`	Account token usage / rate-limit resets (`/usage daily\|weekly\|cumulative`)	Checking quota
`/debug-config`	Config layer and requirements diagnostics	Figuring out where a setting comes from
`/statusline`	Configure the TUI footer status line	Customizing the footer
`/title`	Configure terminal title items	Customizing the window title
`/theme`	Syntax-highlighting theme	Changing colors
`/pets`, `/pet`	Terminal pet	Fun
`/setup-default-sandbox`, `/sandbox-add-read-dir`	Windows-only sandbox helpers	Configuring the Windows sandbox

Renamed/removed commands: older guides mention /approvals (the current docs list /permissions instead) and /rollout (not in the current built-in list). A /recap command and tui.auto_recap setting appear in the 0.153.0 changelog but were not in the captured slash-command table.

/init

What it does

Generates an AGENTS.md scaffold in the current directory — the project-instructions file Codex reads at startup.

When to use it

First session in a repository, or when your existing instructions are stale and you want a fresh scaffold to edit.

Example

> /init

Notes

Edit the generated file rather than accepting it verbatim; specific build/test commands beat generic advice.
For Codex code review in GitHub, add a ## Code Review Rules section to the AGENTS.md closest to the code.
See Codex CLI project instructions (AGENTS.md) for discovery order and limits.

Official source

/new

What it does

Starts a new chat in the same CLI session. /new <name> gives the chat a name.

When to use it

When switching to an unrelated task so the new conversation is not weighed down by old context.

Example

> /new payment-retry-bug

Notes

The previous chat is saved and can be reopened with /resume.
/clear also starts a fresh chat but additionally clears the terminal view.

Official source

https://developers.openai.com/codex/cli/slash-commands

/compact

What it does

Summarizes the visible chat history to free up context tokens.

When to use it

During long sessions, when /status shows little context remaining, or before starting a new subtask in the same chat.

Example

> /compact

Notes

Auto-compaction triggers at model_auto_compact_token_limit; the summarization prompt is customizable via compact_prompt / experimental_compact_prompt_file in config.toml.

Official source

/model

What it does

Opens the model picker to choose the model and reasoning effort for the session.

When to use it

Switch to a stronger model (or higher reasoning effort) for hard problems, or a faster/cheaper one for routine edits.

Example

> /model

Notes

The configured default comes from the model and model_reasoning_effort keys in config.toml; -m/--model overrides it per launch.
Model availability changes quickly — treat specific model names in docs as fast-moving.

Official source

https://developers.openai.com/codex/cli/slash-commands

/review

What it does

Reviews the working tree. Uses the session model unless review_model is set in config.toml.

When to use it

Before committing, to catch bugs in local changes — whether written by you or by Codex.

Example

> /review

Notes

Review does not modify the working tree.
For non-interactive review of branches or commits, use codex review --base main or codex review --commit <SHA> (see Codex CLI terminal commands).

Official source

https://developers.openai.com/codex/cli/slash-commands

/plan

What it does

Switches the active chat into plan mode, optionally with an inline prompt. Codex proposes a plan instead of editing immediately.

When to use it

For multi-file or risky changes where you want to approve an approach before any edits happen.

Example

> /plan Propose a migration plan for this service

Notes

Unavailable while Codex is already working.
plan_mode_reasoning_effort in config.toml sets a plan-mode-specific reasoning override.

Official source

https://developers.openai.com/codex/cli/slash-commands

/diff

What it does

Shows the Git diff of the working tree, including untracked files.

When to use it

After Codex makes edits, to see exactly what changed before you commit or continue.

Example

> /diff

Notes

Untracked files are included, so newly created files show up too.

Official source

https://developers.openai.com/codex/cli/slash-commands

/mention

What it does

Attaches a file to the chat so Codex has its contents as context.

When to use it

When the task concerns a specific file and you want to be sure Codex is looking at it.

Example

> /mention src/payments/retry.ts

Notes

Typing @ in the composer also opens file search/attach.

Official source

https://developers.openai.com/codex/cli/slash-commands

/permissions

What it does

Sets what Codex can do without asking: approval presets (e.g. Auto, Read Only) and custom permission profiles.

When to use it

To raise or lower Codex’s autonomy mid-session — for example dropping to Read Only while you explore, or raising it for a bulk refactor.

Example

> /permissions

Notes

This is the current command for what older docs called /approvals; treat /approvals as renamed.
The equivalent launch-time flags are --sandbox and --ask-for-approval; see Codex CLI permissions, sandbox, and approvals.

Official source

/status

What it does

Shows session configuration and token usage: model, approval policy, writable roots, and context remaining.

When to use it

Any time you are unsure which model, sandbox mode, or writable directories are active — especially after resuming a session.

Example

> /status

Notes

For where a config value came from, use /debug-config.

Official source

https://developers.openai.com/codex/cli/slash-commands

Composer shortcuts

Verified from the official slash-commands page:

Shortcut	Effect
`@`	File search / attach
`!` prefix	Run a local shell command under the current sandbox/approval
Up / Down	Draft history
Ctrl+R	Search prompt history
Tab (while Codex works)	Queue a follow-up for the next turn
Enter (while Codex works)	Inject into the current turn
Esc Esc (empty composer)	Edit the previous message and fork
Ctrl+C	Close the session

Last verified: 2026-09-07

Official sources:

AI coding tools evolve quickly. If you find outdated information, please open an issue or pull request.