Codex CLI PLACEMAT

Keyboard & TUI

Composer

Item	Description
#@	Search workspace files and insert a path into the prompt
#!	Prefix a line to run a shell command under current approval and sandbox settings
#Up / Down	Restore draft history
#Ctrl R	Search prompt history — Enter uses a match, Esc cancels
#Ctrl O	Copy the latest completed Codex output +1
#Ctrl G	Open the composer in $VISUAL, or $EDITOR when unset +1
#Esc Esc	Edit the previous user message and fork the chat from that point +1
#Ctrl C	Close the session (same as /exit)

While Codex is working

Item	Description
#Tab	Queue a follow-up prompt, slash command or shell command for the next turn +2
#Enter	Inject new instructions into the current turn

Appearance & bindings

Item	Description
#/keymap	Inspect, remap and persist TUI shortcut bindings +2
#tui.keymap.<context>.<action>	Bind a TUI action; an empty array unbinds it +2
#/theme	Preview and persist a syntax-highlighting theme +1
#/vim	Toggle Vim normal/insert behaviour in the composer +1
#/raw	Toggle raw scrollback for copy-friendly terminal selection +1
#/statusline	Pick and reorder TUI footer items +1
#/title	Pick and reorder terminal window/tab title items
#codex completion zsh	Generate a shell completion script (bash, zsh, fish, powershell) +2
#--no-alt-screen	Run the TUI inline, preserving terminal scrollback +1

Slash Commands (Core)

Model & behaviour

Item	Description
#/model	Choose the active model, and reasoning effort where available
#/fast	Toggle the current model’s Fast service tier and persist it +1
#/personality	Pick a communication style for responses +1
#/plan	Switch to plan mode, optionally with a prompt +1
#/goal	Set, edit, pause, resume, view or clear a persistent task goal

Permissions & review

Item	Description
#/permissions	Set what Codex may do without asking first
#/approve	Approve one retry of a recent auto-review denial
#/review	Ask Codex to review your working tree
#/diff	Show the Git diff, including files Git is not tracking yet

Session lifecycle

Item	Description
#/new	Start a new chat inside the same CLI session
#/clear	Clear the terminal and start a fresh chat together
#/rename	Rename the current chat
#/resume	Resume a saved chat from your session list
#/fork	Fork the current chat into a new chat
#/side / /btw	Start an ephemeral side chat +1
#/archive	Archive the current session and exit
#/delete	Permanently delete the current session and exit +1
#/quit / /exit	Exit the CLI

Context & output

Item	Description
#/compact	Summarise the visible chat to free tokens
#/mention	Attach a file or folder to the chat
#/copy	Copy the latest completed Codex output
#/ide	Include open files, selection and other IDE context
#/app	Continue the current session in the ChatGPT desktop app (macOS, Windows)
#/status	Show session configuration and token usage +1

Background terminals

Item	Description
#/ps	Show background terminals and their recent output +2
#/stop	Stop all background terminals started by this session

Slash Commands (Tools)

Extensibility

Item	Description
#/skills	Browse and use local skills
#/plugins	Browse installed and discoverable plugins +1
#/hooks	View, trust and manage lifecycle hooks +1
#/mcp	List configured MCP tools +1
#/apps	Browse apps (connectors) and insert them into your prompt +1
#/agent / /subagents	Switch the active agent thread
#/memories	Turn memory injection or memory generation on or off
#/init	Generate an AGENTS.md scaffold in the current directory
#/import	Import Claude Code or Cursor setup, projects and chats
#/experimental	Toggle experimental features such as network proxy or prevent-sleep

Sandbox helpers (Windows)

Item	Description
#/setup-default-sandbox	Set up the elevated agent sandbox +1
#/sandbox-add-read-dir	Grant sandbox read access to an extra absolute directory

Account & diagnostics

Item	Description
#/usage	View account token usage or a rate-limit reset
#/debug-config	Print config layer and requirements diagnostics
#/feedback	Send logs to the Codex maintainers
#/logout	Sign out of Codex
#/pets / /pet	Choose or hide an ambient terminal pet +1

CLI & Subcommands

Global flags

Item	Description
#-m, --model <MODEL>	Model the agent should use
#-c, --config <key=value>	Override one config value for this invocation +2
#-p, --profile <NAME>	Layer $CODEX_HOME/<name>.config.toml over the base config
#-s, --sandbox <MODE>	Sandbox policy: read-only, workspace-write, danger-full-access
#-a, --ask-for-approval <POLICY>	When to require human approval: on-request or never
#-C, --cd <DIR>	Use this directory as the agent’s working root
#--add-dir <DIR>	Additional directories writable alongside the primary workspace +1
#--worktree	Run the session in a new managed Git worktree
#-i, --image <FILE>	Attach one or more images to the initial prompt
#--search	Enable live web search for the model (no per-call approval)
#--oss	Use an open-source provider
#--local-provider <PROVIDER>	Pick the local provider with --oss: lmstudio or ollama
#--enable <FEATURE> / --disable <FEATURE>	Enable or disable a feature flag; both are repeatable +1
#--strict-config	Fail when config.toml has keys this version does not recognise
#--approve-for-me	Route approval requests through automatic review using the workspace-write sandbox
#--dangerously-bypass-approvals-and-sandbox	Skip every prompt and run unsandboxed +1
#--dangerously-bypass-hook-trust	Run enabled hooks without persisted hook trust for this invocation +1
#--remote <ADDR>	Connect the TUI to a remote app server +1
#--remote-auth-token-env <ENV_VAR>	Name of the env var holding the remote app-server bearer token
#-V, --version / -h, --help	Print the version

Sessions & threads

Item	Description
#codex resume	Resume a previous interactive session; picker by default +2
#codex fork	Fork a previous session into a new one; picker by default +1
#codex queue	Queue a message for an existing session
#codex archive	Archive a saved session by id or name
#codex unarchive	Restore an archived session
#codex delete	Permanently delete a saved session by id or name
#codex agents	Browse all agent sessions on the shared local app-server daemon
#codex migrate-rollouts	Inspect or migrate legacy local sessions to paginated thread history

Run & review

Item	Description
#codex exec / codex e	Run Codex non-interactively +3
#codex review	Run a code review non-interactively +1
#codex apply / codex a	Apply the agent’s latest diff to your working tree with git apply

Auth

Item	Description
#codex login	Authenticate with a ChatGPT account, API key or access token +2
#codex logout	Remove stored authentication credentials

MCP & plugins

Item	Description
#codex mcp	Manage external MCP servers +1
#codex plugin	Manage Codex plugins +2

Sandbox & policy

Item	Description
#codex sandbox	Run a command under the same sandbox policy Codex uses internally +1
#codex execpolicy check	Validate execpolicy rule files against a command +2

Diagnostics & maintenance

Item	Description
#codex doctor	Diagnose the local installation, config, auth and runtime health
#codex features	Inspect feature flags +1
#codex debug	Debugging tools +3
#codex update	Update the CLI when the installed release supports self-update
#codex completion	Generate a shell completion script
#codex cloud	Browse Codex Cloud tasks and apply changes locally +1
#codex app-server	Run the app server or related tooling +2
#codex remote-control	Manage the app-server daemon with remote control enabled +1
#codex exec-server	Run the standalone exec-server service +1
#codex mcp-server	Removed in v0.154.0 — use the Codex app server instead +1

Settings (config.toml)

Files & precedence

Item	Description
#~/.codex/config.toml	User-level configuration; the root is $CODEX_HOME +2
#$CODEX_HOME/<name>.config.toml	A config profile layer, selected with --profile <name>
#requirements.toml	Administrator-managed policy layer +1

Model

Item	Description
#model	Model to use
#review_model	Model override used by /review; defaults to the session model
#model_provider	Provider id from model_providers; defaults to openai
#model_reasoning_effort	Reasoning effort: minimal, low, medium, high, xhigh +1
#plan_mode_reasoning_effort	Reasoning effort used while in plan mode
#model_reasoning_summary	Summary detail: auto, concise, detailed, none +1
#model_verbosity	GPT-5 Responses API verbosity override: low, medium, high
#model_context_window	Context window tokens available to the active model
#model_auto_compact_token_limit	Token threshold that triggers automatic history compaction
#personality	Default communication style: none, friendly, pragmatic
#service_tier	Preferred service tier for new turns +1
#oss_provider	Local provider used with --oss: lmstudio or ollama

Approvals & sandbox

Item	Description
#approval_policy	When Codex pauses for approval: on-request or never +2
#approvals_reviewer	Who reviews eligible approval prompts: user or auto_review +1
#sandbox_mode	Sandbox policy: read-only, workspace-write, danger-full-access
#sandbox_workspace_write.writable_roots	Extra directories writable in workspace-write mode
#sandbox_workspace_write.network_access	Allow outbound network access inside the workspace-write sandbox
#sandbox_workspace_write.exclude_slash_tmp / sandbox_workspace_write.exclude_tmpdir_env_var	Exclude /tmp from writable roots
#default_permissions	Named permissions profile applied to sandboxed tool calls +2
#permissions.<name>.extends	Parent profile applied before this one +1
#windows.sandbox	Native Windows sandbox mode: unelevated or elevated
#shell_environment_policy.inherit	Environment inheritance for subprocesses: all, core, none
#allow_login_shell	Allow shell tools to use login-shell semantics; defaults to true
#projects.<path>.trust_level	Trust level for a project path, which gates project-local config and hooks

MCP servers

Item	Description
#mcp_servers.<id>.command / mcp_servers.<id>.args / mcp_servers.<id>.env	Launcher command for a stdio MCP server
#mcp_servers.<id>.url	Endpoint for a streamable HTTP MCP server
#mcp_servers.<id>.auth	Auth fallback for an HTTP MCP server: oauth or chatgpt +2
#mcp_servers.<id>.enabled	Disable a server without removing its configuration
#mcp_servers.<id>.startup_timeout_sec / mcp_servers.<id>.tool_timeout_sec	Override the 10 s startup timeout

Providers

Item	Description
#model_providers.<id>.base_url / model_providers.<id>.name	API base URL for a custom provider
#model_providers.<id>.env_key	Environment variable supplying that provider’s API key
#model_providers.<id>.wire_api	Protocol used by the provider; responses is the only supported value
#openai_base_url	Base URL override for the built-in openai provider
#forced_login_method	Restrict authentication to chatgpt or api

TUI

Item	Description
#tui.notifications	Enable TUI notifications, optionally restricted to event types +2
#tui.theme	Syntax-highlighting theme override, in kebab-case
#tui.status_line	Ordered footer status-line items; null disables the line
#tui.animations	Enable welcome screen, shimmer and spinner animations
#tui.resume_cwd	Directory used when resuming or forking: current or session
#tui.vim_mode_default	Start the composer in Vim normal mode
#tui.raw_output_mode	Start the TUI in raw scrollback mode

Features, hooks & agents

Item	Description
#features.hooks	Enable lifecycle hooks from hooks.json or inline [hooks] +1
#features.apps	Enable app (connector) integrations; on by default +1
#agents.enabled	Enable multi-agent tools; defaults to true
#agents.default_subagent_model / agents.default_subagent_reasoning_effort	Default model for spawned agents
#agents.max_concurrent_threads_per_session	Cap on concurrent spawned-agent threads, excluding the primary thread
#memories.use_memories / memories.generate_memories	Inject existing memories into future sessions
#skills.max_context_tokens	Token budget for the available-skills catalog +1

Housekeeping

Item	Description
#instructions	Reserved — prefer model_instructions_file or AGENTS.md
#developer_instructions	Extra developer instructions injected into the session
#project_doc_max_bytes	Maximum bytes read from AGENTS.md when building project instructions +1
#project_root_markers	Filenames that mark a project root when walking parent directories
#log_dir	Directory for log files; defaults to $CODEX_HOME/log +1
#sqlite_home	Directory for the SQLite-backed state DB used by agent jobs and resumable runs
#file_opener	URI scheme for opening citations: vscode, vscode-insiders, windsurf, cursor, none
#notify	Command invoked for notifications; receives a JSON payload
#check_for_update_on_startup	Check for updates on startup
#hide_agent_reasoning	Suppress reasoning events in the TUI and codex exec
#tool_output_token_limit	Token budget for storing individual tool outputs in history
#history.persistence / history.max_bytes	How prompt history is persisted

Environment Variables

Locations

Item	Description
#CODEX_HOME	Root for config, auth, logs, sessions and skills; defaults to ~/.codex +1
#CODEX_SQLITE_HOME	Where SQLite-backed state is stored; defaults to CODEX_HOME +1

Authentication & network

Item	Description
#CODEX_API_KEY	API key for a non-interactive Codex process +1
#CODEX_ACCESS_TOKEN	ChatGPT or Codex access token for trusted automation +1
#CODEX_CA_CERTIFICATE	PEM CA bundle for corporate TLS interception or private roots +1
#SSL_CERT_FILE	Fallback PEM CA bundle path when CODEX_CA_CERTIFICATE is unset
#OPENAI_FEDERATION_RULE_ID	Selects the workload-identity federation rule
#OPENAI_IDENTITY_TOKEN_FILE	Absolute path to the current OIDC token or SPIFFE JWT-SVID
#OPENAI_WORKLOAD_IDENTITY_CONTEXT	Bounded JSON identifiers for client-reported audit attribution +1

Installer & diagnostics

Item	Description
#CODEX_NON_INTERACTIVE	Skip installer prompts, using their defaults +1
#CODEX_INSTALL_DIR	Where the visible codex command is installed +2
#RUST_LOG	Log filtering and verbosity +2

Skills, Plugins & AGENTS.md

Skills

Item	Description
#/skills	Browse and use local skills from the TUI
#SKILL.md	The file that defines a skill inside its folder
#$CWD/.agents/skills	Repo-scoped skills; Codex scans every directory from the CWD up to the repo root +3
#$HOME/.agents/skills	User-scoped skills, available in every repository
#/etc/codex/skills	Admin-scoped skills shared across a machine or container
#skills.config	Per-skill enablement overrides +1
#~/.codex/prompts	Deprecated custom prompts — use skills instead +1

Plugins & marketplaces

Item	Description
#/plugins	Browse installed and discoverable plugins from the TUI
#codex plugin add / codex plugin remove	Install a plugin from a configured or remote marketplace
#marketplaces.<name>.source_type	Marketplace source kind: git or local +2
#plugins.<plugin>.enabled	Enable or disable a plugin, keyed plugin-name@marketplace-name +2
#plugins.<plugin>.mcp_servers.<server>.enabled_tools	Allow list of tools from a plugin-provided MCP server +2

AGENTS.md

Item	Description
#/init	Generate an AGENTS.md scaffold in the current directory
#AGENTS.md	Project guidance, merged root-down so nearer files win +3
#AGENTS.override.md	Takes precedence over AGENTS.md in the same directory
#project_doc_fallback_filenames	Extra filenames accepted when no AGENTS.md is found in a directory

Subagents

Item	Description
#/agent / /subagents	Inspect or continue work in a spawned subagent thread
#agents.<name>.description	Role guidance shown to Codex when choosing and spawning that agent type +2
#codex agents	Browse all agent sessions on the shared local app-server daemon

Hooks, MCP & Sandbox

Hooks

Item	Description
#/hooks	View configured hooks, trust new or changed ones, disable non-managed hooks
#hooks.json	Hook definitions next to an active config layer +3
#[hooks]	Inline hook tables in config.toml, using the same event schema as hooks.json
#PreToolUse / PostToolUse / PermissionRequest	Lifecycle events you can hook +2
#hooks.<Event>[].hooks[].async	Run a command hook in the background without delaying the trigger +1
#hooks.<Event>[].hooks[].additionalContextLimit	Token threshold before oversized hook context is spilled to disk +1
#hooks.<Event>[].hooks[].commandWindows	Windows-only command override for a command hook +1
#--dangerously-bypass-hook-trust	Run enabled hooks without persisted trust for this invocation +1

MCP

Item	Description
#codex mcp add / codex mcp list / codex mcp get / codex mcp remove	Register an external MCP server
#codex mcp login / codex mcp logout	Authenticate against an MCP server
#mcp_oauth_callback_port / mcp_oauth_callback_url	Port for the MCP OAuth callback
#mcp_optional_startup_grace_ms	Grace period before optional MCP servers are considered failed
#codex app-server	The supported way to expose Codex to another client +1

Sandbox & exec policy

Item	Description
#read-only / workspace-write / danger-full-access	Sandbox modes for model-generated commands +1
#codex sandbox	Run any command under the same policy Codex uses internally +1
#codex execpolicy check --rules	Validate execpolicy rule files against a command before saving them +1
#~/.codex/rules	Conventional location for execpolicy rule files +1