Codex CLI Slash Commands Complete Reference: All 45 Commands in v0.131

codex-clislash-commandstuireferencev0.131keyboard-shortcutsworkflow

Sketchnote diagram for: Codex CLI Slash Commands Complete Reference: All 45 Commands in v0.131

Codex CLI Slash Commands Complete Reference: All 45 Commands in v0.131

Codex CLI v0.131.0 ships with 45 slash commands — more than triple the count from early 20261. Most guides still reference the original dozen. This article catalogues every command available in the TUI as of May 2026, organised by function, with usage patterns that senior developers actually need.

How Slash Commands Work

Type / in the Codex CLI composer to open the command picker. Tab-completion filters the list as you type. Most commands accept inline arguments (e.g. /plan refactor the auth module) and execute immediately2.

Slash commands are distinct from CLI subcommands like codex exec or codex mcp. They operate inside an active interactive session, not from the shell.

Model and Performance Control

These commands govern which model processes your prompts and how fast it responds.

Command Purpose Example
/model Switch the active model mid-session /model gpt-5.4-mini
/fast Toggle the Fast service tier (1.5× speed, higher credit burn) /fast on
/personality Choose a communication style: friendly, pragmatic, or none /personality pragmatic

/fast works with GPT-5.5 (2.5× credit consumption) and GPT-5.4 (2× consumption)3. API key users cannot access Fast mode — it requires a ChatGPT subscription. Use /personality none to strip all personality instructions and reclaim token budget for task context.

# Persist these in ~/.codex/config.toml instead of setting per-session
model = "gpt-5.4-mini"
service_tier = "fast"

[features]
fast_mode = true

Session Management

Commands for controlling conversation flow, branching, and resuming prior work.

Command Purpose
/clear Wipe the terminal and start a fresh chat within the same CLI process
/new Start a new conversation thread without leaving the CLI
/resume Reload a previous saved conversation from the session list
/fork Clone the current conversation into a new thread, preserving the original
/side Open an ephemeral side conversation without derailing the main thread
/exit Close the interactive session
/quit Alias for /exit

Branching Patterns

/fork creates a persistent branch — the original transcript survives, and you can /resume either branch later. Use it when you want to try an alternative architectural approach without losing progress4.

/side creates a throwaway branch. Ask a quick question, then return to the main thread with no transcript pollution. Useful for ad-hoc queries mid-refactor:

/side what's the maximum connection pool size for pg 16?

The side conversation’s output stays visible for reference but does not enter the main thread’s context window.

Workflow and Mode Switching

These commands change how Codex interacts with your codebase.

Command Purpose
/plan Switch to plan mode — Codex reads files and proposes a plan without writing anything
/goal Set a persistent objective that survives across turns (experimental)
/review Ask Codex to review your working tree, focusing on behaviour changes and missing tests
/diff Display the Git diff, including staged changes, unstaged changes, and untracked files
/compact Summarise earlier conversation to free context tokens 
flowchart LR
    A["/plan"] --> B["Codex proposes<br/>implementation plan"]
    B --> C{"Approve?"}
    C -- Yes --> D["Shift+Tab to<br/>Execute mode"]
    D --> E["Codex implements"]
    E --> F["/review"]
    F --> G["/diff"]
    C -- No --> H["Refine prompt"]
    H --> B

Plan-Execute-Review Cycle

The canonical workflow is /plan → review the proposal → Shift+Tab to switch to Execute mode → /review to validate → /diff to inspect changes5. This keeps the agent from racing ahead on an approach you haven’t sanctioned.

Context Compaction

/compact is essential for long sessions. When context pressure rises, Codex replaces earlier turns with a compressed summary. The model preserves task-critical details — file paths, decisions made, error signatures — whilst discarding conversational boilerplate6.

Persistent Goals

/goal sets an objective that persists across turns:

/goal raise test coverage in src/auth to 90% without mocking external calls
/goal          # view the current goal
/goal pause    # temporarily suspend
/goal resume   # reactivate
/goal clear    # remove entirely

This is an experimental feature requiring the goals feature flag7.

Permissions and Security

Commands for adjusting the agent’s autonomy boundary during a session.

Command Purpose
/permissions Switch approval mode: auto-approve, on-request, or manual
/approve Retry a specific action that the auto-reviewer denied
/hooks Review, trust, or disable lifecycle hooks

/permissions maps to the --ask-for-approval CLI flag. Tightening mid-session (e.g. from never to on-request) takes effect immediately without restarting8.

/approve is narrow by design: it retries only the most recently denied action, not a blanket override. This prevents accidental escalation when the auto-reviewer blocks a sandbox boundary crossing or a side-effecting MCP call9.

Context and File Attachment

Commands for injecting additional context into the conversation.

Command Purpose
/ide Include open files, current selection, and other IDE context
/mention Attach a specific file to the conversation
/memories Configure whether threads use or generate persistent memories

/mention is equivalent to the @ mention syntax in the composer. Both support tab-completion across files, directories, plugins, and skills as of v0.131.0’s unified mentions system10.

/memories controls two dimensions independently:

/memories generate off    # this thread won't create new memories
/memories use on          # this thread will consume existing memories

Agent and Extension Management

Commands for working with subagents, plugins, skills, apps, and MCP tools.

Command Purpose
/agent Switch the active agent thread when running parallel subagents
/skills Browse and invoke task-specific skills
/apps Browse connectors (apps) and insert them as $app-slug mentions
/plugins Browse installed and marketplace plugins
/mcp List configured MCP tools and their status

Subagent Switching

When you’ve delegated work to subagents, /agent lets you switch focus between threads. Each thread maintains its own context, approval state, and model configuration11.

Skill Invocation

/skills opens a picker showing all available skills. Select one and Codex loads its SKILL.md into context. You can also invoke skills implicitly by starting a prompt with $skill-name12.

TUI Customisation

Commands for tailoring the terminal interface to your preferences.

Command Purpose
/vim Toggle Vim modal editing in the composer
/raw Toggle raw scrollback mode for cleaner copy-paste
/keymap Remap TUI keyboard shortcuts
/statusline Configure which fields appear in the TUI footer
/title Customise terminal window or tab title fields
/theme Choose a syntax-highlighting theme with live preview

/vim was added in v0.131.0 and supports normal, insert, and visual modes within the composer. You can set it as the default:

[tui]
default_mode = "vim"

/raw is useful when you need to copy a large block of output without TUI chrome interfering. Toggle it on, select and copy, then toggle it off to return to the rich display13.

Reasoning Controls

Whilst not slash commands, the keyboard shortcuts Alt+, (lower reasoning effort) and Alt+. (raise reasoning effort) pair naturally with /model for fine-grained control over response quality and speed14.

Diagnostics and System

Commands for inspecting configuration, reporting issues, and managing authentication.

Command Purpose
/status Display session configuration, model, approval mode, and token usage
/debug-config Print the full configuration layer stack and requirements diagnostics
/experimental Toggle experimental feature flags
/feedback Submit session logs to the Codex maintainers
/init Generate an AGENTS.md scaffold in the current directory
/logout Clear stored authentication credentials

/debug-config is the session equivalent of codex doctor. It shows which configuration layers are active, their precedence, any requirements.toml constraints from enterprise governance, and flag overrides. Invaluable when a setting isn’t behaving as expected15.

Background Process Management

Commands for monitoring and controlling background work.

Command Purpose
/ps Show experimental background terminals and their status
/stop Cancel all running background terminals
/copy Copy the latest completed Codex output to the clipboard

/ps lists any background processes spawned by Codex during the session — useful when the agent kicks off a long-running build or test suite in the background and you need to check progress16.

Quick Reference Card

For rapid lookup, here are all 45 commands in alphabetical order:

/agent       /approve     /apps        /clear       /compact
/copy        /debug-config /diff       /exit        /experimental
/fast        /feedback    /fork        /goal        /hooks
/ide         /init        /keymap      /logout      /mcp
/memories    /mention     /model       /new         /permissions
/personality /plan        /plugins     /ps          /quit
/raw         /resume      /review      /sandbox-add-read-dir
/side        /skills      /status      /statusline  /stop
/theme       /title       /vim

Note: /sandbox-add-read-dir is Windows-only and grants read access to additional directories at runtime17.

Building Custom Slash Commands

Codex does not currently support user-defined slash commands directly. However, you can achieve similar effects through:

  1. Skills — Package a workflow as a skill with a SKILL.md file, then invoke it via /skills or $skill-name12
  2. Hooks — Attach scripts to lifecycle events (pre-tool-use, post-tool-use, on-turn-end) for automated validation18
  3. Plugins — Bundle skills, hooks, and MCP servers together and distribute via the marketplace19

What Changed in v0.131.0

The following commands were added or significantly enhanced in the v0.131.0 release cycle1:

  • /vim — New. Modal Vim editing in the composer with configurable default mode
  • /approve — New. Retry auto-review denials without restarting
  • /agent — Enhanced. Now supports the multi-environment session architecture
  • /apps — Enhanced. Backed by app-server plugin metadata with unified search
  • /statusline — New. Interactive footer configuration with blended token usage display
  • /title — New. Terminal title customisation
  • /debug-config — New. Configuration layer diagnostics

The unified @ mentions system also changed how /mention works — it now searches files, directories, plugins, and skills through a single picker10.

Citations

  1. Codex CLI v0.131.0 Release Notes — GitHub, May 2026 ↩ ↩2

  2. Slash Commands in Codex CLI — OpenAI Developer Documentation ↩

  3. Codex Speed Configuration — OpenAI Developer Documentation ↩

  4. Features — Codex CLI — OpenAI Developer Documentation ↩

  5. Best Practices — Codex — OpenAI Developer Documentation ↩

  6. Codex Prompting Guide — OpenAI Cookbook ↩

  7. Codex /goal Command — OpenAI Developer Documentation ↩

  8. Agent Approvals and Security — OpenAI Developer Documentation ↩

  9. Auto-Review Documentation — OpenAI Developer Documentation, May 2026 ↩

  10. Codex Changelog — v0.131.0 — OpenAI, May 18, 2026 ↩ ↩2

  11. Subagents — Codex — OpenAI Developer Documentation ↩

  12. Agent Skills — Codex — OpenAI Developer Documentation ↩ ↩2

  13. Codex CLI Features — Raw Scrollback — OpenAI Developer Documentation ↩

  14. Codex Changelog — TUI Reasoning Controls — OpenAI, May 2026 ↩

  15. Config Reference — Codex — OpenAI Developer Documentation ↩

  16. Codex CLI Features — Background Terminals — OpenAI Developer Documentation ↩

  17. Command Line Options — Codex CLI — OpenAI Developer Documentation ↩

  18. Hooks — Codex — OpenAI Developer Documentation ↩

  19. Plugins — Codex — OpenAI Developer Documentation ↩