Codex CLIチートシート：開発者向けコマンド完全ガイド

Printable single-page A4 JPEG reference for developersLast updated: 09/05/2026Concise reference for the Codex CLI: terminal UI, codex exec, local config, MCP, skills, subagents, hooks, rules, and automation.🚀 Start HereCommandUsecodexOpen the interactive terminal UI in the current directorycodex "prompt"Start an interactive session with an initial instructioncodex exec "prompt" / codex e "prompt"Run Codex non-interactively for scripts or CIcodex resumePick and resume a previous interactive sessioncodex resume --lastResume the latest session for this directorycodex fork --lastFork the latest session into a new threadcodex loginSign in with ChatGPT OAuthcodex login --device-authSign in on remote/headless machinescodex login --with-api-keyStore API-key auth from stdin, e.g. pipe OPENAI_API_KEYcodex login statusCheck auth; exits 0 when logged incodex logoutRemove cached credentialscodex updateCheck/apply a CLI update when supportedcodex completion <shell>Generate completions for bash, zsh, fish, power-shell, or elvish🖥️ Interactive TUIInputAction@Fuzzy file search; insert a path into the prompt/commandRun a slash command!cmdRun a local shell command under current approvals/sandboxCtrl+C or /exitExit the sessionCtrl+LClear/redraw screen without starting a new conversationCtrl+O or /copyCopy latest completed Codex outputCtrl+RSearch prompt historyCtrl+GOpen prompt in $VISUAL or $EDITORTab while Codex runsQueue follow-up prompt, slash command, or ! commandEnter while Codex runsInject instructions into the current turnEsc Esc on empty composerEdit previous user message; keep pressing Esc to go further back⚡ Slash CommandsUse /help in the TUI for the exact command list in your installed version.CommandUse/statusInspect current session state, workspace, and session id/permissionsSwitch approval/sandbox mode/clearClear current conversation context/compactSummarize long context; Codex can also auto-compact/resumeResume a saved conversation from the TUI/forkBranch a new thread from the current transcript/sideStart a side conversation/modelSwitch model and reasoning choices/fast on, /fast off, /fast statusToggle or inspect Fast mode/planEnter Plan mode before implementation/review [focus]Run local code review presets/mcpShow active MCP servers/agentSwitch/inspect active subagent threads/appsUse enabled ChatGPT app/connectors from the CLI/pluginsBrowse, install, uninstall, or toggle plugins/skills or $skill-nameBrowse or explicitly invoke skills/themePreview/save syntax theme to tui.theme/mentionAttach a specific file as context/experimentalToggle experimental features into config/personalityChange supported-model communication style/feedbackSend feedback, unless disabled by config⚙️ Global FlagsFlagUse--model <model>, -mOverride configured model, e.g. gpt-5.5 or gpt-5.4--profile <name>, -pLoad [profiles.<name>] from config.toml--config key=value, -c key=valueOne-off config override; value is parsed as TOML--cd <path>, -C <path>Set workspace root before Codex starts--add-dir <path>Grant extra writable roots; repeatable--image <path[,path...]>, -iAttach image files to the first prompt--sandbox <mode>, -sread-only, workspace-write, or danger-full-access--ask-for-approval <mode>, -auntrusted, on-request, or never--searchUse live web search instead of cached search--ossUse local OSS provider; requires Ollama or configured local provider--enable <feature> / --disable <feature>Force feature flag on/off for this run--no-alt-screenDisable alternate-screen TUI rendering--remote ws://host:portConnect TUI to a remote app server--remote-auth-token-env ENVRead remote WebSocket bearer token from env var--yoloAlias for --dangerously-bypass-approvals-and-sandbox ⚠️🤖 codex execCommand / FlagUsecodex exec "prompt"Run once; progress on stderr, final answer on stdoutcodex exec -Read the full prompt from stdincodex exec "instruction" with piped stdinTreat command output as extra contextcodex exec resume --last "prompt"Continue latest non-interactive sessioncodex exec resume <SESSION_ID> "prompt"Resume a specific exec session--jsonEmit JSONL events: turns, items, tools, plans, errors--output-last-message <file>, -o <file>Save final assistant message--output-schema <file>Require final output matching JSON Schema--ephemeralDo not persist session rollout files--ignore-user-configSkip $CODEX_HOME/config.toml--ignore-rulesSkip user/project .rules files--skip-git-repo-checkAllow running outside a Git repo--color always, never, or autoControl ANSI colorCODEX_API_KEY=... codex exec ...API-key auth for a single exec workflow🔐 Approvals & SandboxSettingMeaning--sandbox read-onlyRead files only; no edits or commands without approval--sandbox workspace-writeRead/edit inside workspace and --add-dir roots--sandbox danger-full-accessNo filesystem sandbox; use only in isolated environments--ask-for-approval untrustedPrompt before untrusted commands--ask-for-approval on-requestPrompt when Codex requests escalation--ask-for-approval neverNever prompt; best for non-interactive, tightly sandboxed runsapprovals_reviewer = "auto_review"Route eligible approvals through automatic reviewapproval_policy = { granular = { ... } }Allow/deny categories such as sandbox, rules, MCP, skillsdefault_permissions = ":workspace"Use built-in permission profiledefault_permissions = "custom"Use [permissions.custom] profileRecommended default: codex --sandbox workspace-write --ask-for-approval on-request.Protected in workspace-write mode: .git/, .codex/, and .agents/ are read-only when present. Use --add-dir or named permission profiles before reaching for full access.🧩 Config Files & KeysPath / KeyUse~/.codex/config.tomlUser config.codex/config.tomlProject config; loaded only for trusted projects/etc/codex/config.tomlUnix system config, if presentCODEX_HOMEOverrides ~/.codex state/config root~/.codex/auth.jsonFile credential cache; treat like a password~/.codex/history.jsonlLocal history when enabled~/.codex/log/Local logs such as codex-tui.logrequirements.tomlAdmin-enforced constraintsPrecedence: CLI flags/-c > profile > project .codex/config.toml > user config > system config > defaults.Common KeyValues / UsemodelDefault modelreview_modelModel override for /reviewmodel_providerProvider id, default openaiopenai_base_urlProxy/data-residency base URL for built-in OpenAI providermodel_reasoning_effortminimal, low, medium, high, xhighplan_mode_reasoning_effortPlan-mode-specific effort overridemodel_reasoning_summaryauto, concise, detailed, nonemodel_verbositylow, medium, highpersonalitynone, friendly, pragmaticservice_tierflex or fastweb_searchcached, live, or disabledproject_root_markersRoot detection markers, e.g. [".git", ".hg"]project_doc_max_bytesMax bytes read from AGENTS.md guidanceproject_doc_fallback_filenamesExtra instruction filenamesfile_openervscode, cursor, windsurf, vscode-insiders, nonehistory.persistencesave-all or noneanalytics.enabledLocal analytics togglefeedback.enabledEnable /feedbackcli_auth_credentials_storefile, keyring, or auto🧠 Models, Speed, SearchFeatureHowRecommended modelUse gpt-5.5 when available; otherwise gpt-5.4Switch model in TUI/modelSet model for runcodex --model gpt-5.5Fast mode/fast on, /fast off, /fast statusPersist fast modeservice_tier = "fast" and [features].fast_mode = trueCodex-Sparkgpt-5.3-codex-spark, fast research-preview model for Pro usersCached searchDefault local web search modeLive search--search or web_search = "live"Disable searchweb_search = "disabled"📁 AGENTS.md GuidanceLocationPurpose~/.codex/AGENTS.mdGlobal personal guidance~/.codex/AGENTS.override.mdTemporary global override<repo>/AGENTS.mdRepo-level shared instructions<subdir>/AGENTS.override.mdLocal override for a subtreeproject_doc_fallback_filenamesTreat other files as instruction filesDiscovery order: global first, then project root down to current directory. In each directory Codex prefers AGENTS.override.md, then AGENTS.md, then fallback filenames. Later, closer files win.Use /init to scaffold a starter AGENTS.md.🔌 MCPCommand / ConfigUse/mcpView active MCP servers in TUIcodex mcp list [--json]List configured serverscodex mcp get <name> [--json]Show one servercodex mcp add <name> -- <command...>Add stdio servercodex mcp add <name> --url https://...Add streamable HTTP servercodex mcp remove <name>Remove servercodex mcp login <name>OAuth login for supported HTTP serverscodex mcp logout <name>Remove stored MCP OAuth credentialscodex mcp-serverRun Codex itself as an MCP server over stdiomcp_servers.<id> KeyUsecommand, args, env, env_vars, cwdStdio server launchurl, bearer_token_env_varStreamable HTTP serverhttp_headers, env_http_headersStatic/env-backed HTTP headersenabled = falseDisable without deletingrequired = trueFail startup/resume if server cannot initializestartup_timeout_sec, tool_timeout_secServer/tool timeoutsenabled_tools, disabled_toolsTool allow/deny listsmcp_oauth_callback_port, mcp_oauth_callback_urlOAuth callback overrides🛠️ Skills & PluginsSurfaceUse$skill-nameExplicitly invoke a skill/skillsBrowse available skills$skill-creatorScaffold a new skill$skill-installer <name>Install curated/local skills[[skills.config]] path = ".../SKILL.md"; enabled = falseDisable a skill/pluginsBrowse/install/uninstall/toggle plugins in CLIcodex plugin marketplace add <source>Add plugin marketplacecodex plugin marketplace upgrade [name]Refresh marketplace(s)codex plugin marketplace remove <name>Remove marketplaceSkill LocationScope.agents/skills/<name>/SKILL.mdRepo/team skill, searched from CWD up to repo root$HOME/.agents/skills/<name>/SKILL.mdUser skill/etc/codex/skills/<name>/SKILL.mdAdmin/system skillBundled skillsBuilt into CodexSKILL.md requires name and description. Optional folders: scripts/, references/, assets/, and agents/openai.yaml. Package reusable skills as plugins when you want distribution plus optional apps/MCP config.👥 SubagentsConceptUseBuilt-insdefault, worker, explorer/agentSwitch between active agent threadsExplicit promptCodex only spawns subagents when you askagents.max_threadsConcurrent open agent cap; default 6agents.max_depthSpawn nesting depth; default 1agents.job_max_runtime_secondsDefault timeout for CSV fan-out workersspawn_agents_on_csvExperimental batch fan-out from CSVCustom agents live in ~/.codex/agents/*.toml or .codex/agents/*.toml and require:name = "reviewer"

description = "PR reviewer focused on correctness, security, and missing tests."

developer_instructions = "Review code like an owner. Lead with real risks."

※ 著作権に配慮し、引用は冒頭3段落までです。続きは元記事をご覧ください。