Codex CLI TUI Shortcuts and Slash Commands: The Complete Reference

Updated

referenceconfig-tomlcodex-cli

Sketchnote diagram for: Codex CLI TUI Shortcuts and Slash Commands: The Complete Reference

Codex CLI TUI Shortcuts and Slash Commands: The Complete Reference

Codex CLI’s full-screen terminal UI (TUI) is where most interactive work happens. Beneath the chat-style composer sits a dense set of keyboard shortcuts, input modifiers and slash commands that let you control models, review diffs, manage sessions and configure the agent without leaving the terminal1. This reference catalogues every shortcut and slash command available as of May 2026 (v0.135.0), organised by function. Updated with /plugins, codex doctor, Vim mode and the current model names.

The composer: input shortcuts and modifiers

The composer is the text-entry area at the bottom of the TUI. Beyond plain text, it supports several input modifiers and keyboard shortcuts that speed up interaction.

Keyboard shortcuts

Shortcut Action
Enter Send the current prompt. During agent execution, injects new instructions into the running turn (steer mode)2
Tab During agent execution, queues a follow-up prompt for the next turn rather than interrupting the current one2
Ctrl+G Opens the current prompt in your external editor, as defined by VISUAL (falls back to EDITOR)3
Ctrl+L Clears the terminal screen without resetting the conversation context3
Ctrl+C Cancels the current operation; press twice to quit the session4
Ctrl+D Exits Codex CLI; press twice to force quit4
Esc, Esc When the composer is empty, double-pressing Escape edits your previous message. Continue pressing to walk back through the transcript3
Up / Down Navigate through draft history in the composer. Codex restores prior draft text and image placeholders3

Since v0.135.0, the TUI also supports Vim-style text-object editing when Vim mode is enabled, including word, line-end behaviour and a configurable interrupt-turn binding.

Input modifiers

Three prefix characters transform composer input before it reaches the model:

Prefix Purpose Example
@ Fuzzy file search over the workspace root. Press Tab or Enter to accept a match and attach the file to the conversation3 @src/main.rs
! Executes a local shell command outside Codex’s sandbox. Output is displayed but not sent to the model3 !git log --oneline -5
-i / –image Attach images via CLI flags at launch, or paste screenshots directly into the composer during a session5 codex -i mockup.png "Implement this design" 
flowchart LR
    A[User types in composer] --> B{First character?}
    B -->|@| C[Fuzzy file search]
    B -->|!| D[Local shell exec]
    B -->|/ | E[Slash command popup]
    B -->|text| F[Send to model]
    C -->|Tab/Enter| G[Attach file to context]
    D --> H[Display output locally]
    E --> I[Execute command]
    F -->|Enter| J[Model processes prompt]

Slash commands: complete reference

Type / in the composer to open the slash-command popup. Commands are grouped below by function1.

Session management

Command Description
/clear Clears the terminal, resets the visible transcript and starts a fresh chat within the same CLI session1
/new Starts a new conversation inside the same session without leaving the terminal1
/resume Resumes a saved conversation from your session list1
/fork Forks the current conversation into a new thread, preserving the original1
/compact Summarises the visible conversation to free tokens while retaining key context1
/copy Copies the latest completed Codex output to the system clipboard1
/exit Exits the CLI session1
/quit Alias for /exit1

Model and configuration

Command Description
/model Opens a picker to choose the active model (e.g. gpt-5.5, gpt-5.4, gpt-5.4-mini, gpt-5.3-codex, gpt-5.3-codex-spark) and adjust reasoning effort16
/fast Toggles fast mode for GPT-5.4. Accepts /fast on, /fast off or /fast status1
/permissions Opens an approval-preset picker: choose between named permission profiles mid-session. Since v0.135.0, displays configured custom profiles1
/personality Changes how Codex communicates (tone, verbosity) without altering system instructions1
/statusline Configures which fields appear in the TUI footer status bar and their order1
/theme Previews and saves syntax-highlighting colour schemes for code blocks in the TUI3
/experimental Toggles experimental features such as subagents on or off1
/debug-config Prints configuration layer order and policy sources for debugging precedence issues1

Code and project tools

Command Description
/diff Shows the Git diff of the current working tree, including untracked files1
/review Launches a code review of your working tree changes. Supports presets: against a base branch, uncommitted changes or specific commits13
/plan Switches to plan mode, optionally with a prompt. The agent produces an execution plan without making changes1
/init Generates an AGENTS.md scaffold in the current directory, capturing persistent repository instructions1
/mention Attaches a file to the conversation for the model to inspect1

Agent and tool management

Command Description
/agent Switches the active agent thread when working with spawned subagents1
/ps Shows experimental background terminals and their recent output1
/mcp Lists all configured Model Context Protocol tools available in the session1
/apps Browses available apps (connectors) and inserts them into your prompt1
/plugins Opens the plugin browser to search, install and manage marketplace plugins1
/reload-plugins Reloads all installed plugins without restarting the session1

Platform and diagnostics

Command Description
/status Displays the active model, approval policy, writable roots and current token usage1
/feedback Sends session logs to the Codex maintainers for diagnostics and bug reports1
/logout Signs out and clears stored credentials, useful on shared machines1
/sandbox-add-read-dir Grants the sandbox read access to additional directories (Windows-specific)1

Steer mode versus plan mode: Enter and Tab mechanics

Understanding the difference between Enter and Tab during agent execution is essential for effective steering2:

sequenceDiagram
    participant U as User
    participant A as Agent (running)

    Note over A: Agent is executing a task...
    U->>A: Press Enter + type correction
    Note over A: Injects immediately into current turn
    A->>A: Adjusts behaviour mid-execution

    Note over A: Agent continues working...
    U->>A: Press Tab + type follow-up
    Note over A: Queued for next turn
    A->>A: Finishes current task
    A->>U: Reads queued prompt
    A->>A: Begins follow-up work
  • Enter (steer mode): sends your message immediately, interrupting the agent’s current work. Use this for urgent corrections, ‘stop, don’t delete that file’ or ‘use the staging database instead’2.
  • Tab (queue mode): holds your prompt until the agent finishes its current turn. Use this for follow-up tasks, ‘after that, run the test suite’ or ‘then update the changelog’2.

CLI subcommands for non-interactive use

These subcommands complement the interactive session and share the same global flags7:

Subcommand Purpose
codex exec / codex e Runs Codex non-interactively with a prompt; ideal for CI/CD pipelines7
codex review Non-interactive code review against the working tree7
codex resume Resumes a previous interactive session from the command line7
codex fork Forks a previous session into a new one7
codex apply / codex a Applies the latest diff from a Cloud task as git apply7
codex cloud Browses and manages Cloud tasks from the terminal (experimental)7
codex mcp Manages MCP server configurations7
codex doctor Runs environment, Git, terminal, app-server and thread diagnostics for troubleshooting (enhanced in v0.135.0)7
codex remote-control Starts or manages the remote control daemon for headless/remote sessions7
codex completion bash\|zsh\|fish Generates shell completion scripts7
codex sandbox Runs arbitrary commands under the sandbox policy for testing7

Global flags worth knowing

These flags apply to the base codex command and propagate to subcommands7:

# Override model for this session
codex -m gpt-5.5 "Refactor the auth module"

# Inline config overrides
codex -c model_provider="oss" -c sandbox="workspace-write"

# Attach images at launch
codex -i screenshot.png -i wireframe.png "Implement this UI"

# Load a named profile
codex -p fast-review

# Set approval policy
codex -a never "Run the full test suite"

# Grant additional directory access
codex --add-dir /var/log "Search recent error logs"

Platform differences

Feature macOS Linux Windows (WSL)
Sandbox engine Seatbelt8 Landlock + seccomp8 Restricted tokens + ACLs8
/sandbox-add-read-dir Not needed Not needed Available1
codex app subcommand Available Not available Not available7
Ctrl+G editor Respects VISUAL / EDITOR Respects VISUAL / EDITOR Respects VISUAL / EDITOR3
Shell completions bash, zsh, fish bash, zsh, fish bash, zsh (via WSL)7

Quick-reference card

The shortcuts worth committing to muscle memory:

Action Shortcut
Attach a file @filename + Tab
Run a shell command !command
Review changes before committing /diff then /review
Switch model mid-session /model
Free up context window /compact
Copy last response /copy
Open long prompt in editor Ctrl+G
Correct agent mid-task Enter + message
Queue follow-up work Tab + message
Check token usage /status

Citations

  1. Slash commands in Codex CLI — OpenAI Developers ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15 ↩16 ↩17 ↩18 ↩19 ↩20 ↩21 ↩22 ↩23 ↩24 ↩25 ↩26 ↩27 ↩28 ↩29 ↩30 ↩31 ↩32 ↩33

  2. Features — Codex CLI — OpenAI Developers ↩ ↩2 ↩3 ↩4 ↩5

  3. OpenAI Codex CLI Cheat Sheet — Shortcuts and Commands — ComputingForGeeks ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9

  4. Codex CLI Cheat Sheet 2026 — Commands and Options — Toolsbase ↩ ↩2

  5. CLI — Codex — OpenAI Developers ↩

  6. Command line options — Codex CLI — OpenAI Developers ↩

  7. Command line options — Codex CLI — OpenAI Developers ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8 ↩9 ↩10 ↩11 ↩12 ↩13 ↩14 ↩15

  8. GitHub — openai/codex: Lightweight coding agent that runs in your terminal ↩ ↩2 ↩3