{"slug":"codex-cli-reference","title":"OpenAI Codex CLI: Commands, Sandbox Modes, and Workflow Patterns","tags":["codex","openai","cli","agents","coding-assistant"],"agent_summary":"OpenAI Codex CLI quick reference — installation, authentication, sandbox modes (suggest/auto/full), slash commands, workflow patterns, and comparison with Claude Code equivalents.","trigger_phrases":["Codex CLI","OpenAI Codex","codex command","codex --yolo","Codex sandbox","Codex slash command"],"runnable":false,"markdown":"\n## Overview\n\nOpenAI Codex CLI is a terminal-based AI coding assistant. It runs in a project directory and can read files, write code, run commands, and iterate based on test output.\n\n## Installation\n\n```bash\nnpm install -g @openai/codex\n\n# Verify\ncodex --version\n```\n\nRequires Node.js 18+.\n\n## Authentication\n\n| Method | When |\n|--------|------|\n| ChatGPT login | Any ChatGPT subscription — browser OAuth redirect |\n| API key | No subscription, or programmatic access |\n\nPlans: Go ($8/mo), Plus ($20/mo), Pro ($100–$200/mo). Pro recommended for fast mode and high reasoning tasks.\n\n## Commands Quick Reference\n\n| Codex | Claude Code Equivalent | Description |\n|-------|----------------------|-------------|\n| `codex` | `claude` | Launch agent in current directory |\n| `codex --yolo` | `claude --dangerously-skip-permissions` | Full access, no approval prompts |\n| `codex --model o3` | `--model claude-opus-4-6` | Specify model |\n| `codex --quiet` | n/a | Minimal output, automation-friendly |\n\n## Sandbox Modes\n\nCodex runs in one of three modes, selected at startup:\n\n| Mode | Behavior | When to Use |\n|------|----------|-------------|\n| **suggest** (default) | Proposes changes, you approve each | Unfamiliar codebase, learning |\n| **auto** | Auto-applies low-risk changes, asks on risky ones | Day-to-day development |\n| **full** (`--yolo`) | No approvals, runs everything | Trusted projects, CI environments |\n\n## Slash Commands\n\nCodex supports slash commands in the interactive session:\n\n```\n/clear — clear conversation context\n/history — show conversation history\n/diff — show uncommitted changes\n/files — list files in context\n/model — switch model mid-session\n/quit — exit\n```\n\n## Workflow Patterns\n\n### Pattern 1: Plan First, Then Build\n\n```bash\ncodex \"Create a plan for adding Stripe subscriptions to this Next.js app. \\\n List all files that need to change. Don't write code yet.\"\n\n# Review the plan\ncodex \"Execute the plan. Write all the code.\"\n```\n\n### Pattern 2: Test-Driven Development\n\n```bash\ncodex \"Write tests for the UserService class first. \\\n Then implement the class to make them pass.\"\n```\n\n### Pattern 3: Iterative Bug Fix\n\n```bash\ncodex \"Run the test suite and fix any failing tests. \\\n Keep running until all tests pass.\"\n```\n\n### Pattern 4: Architecture Review\n\n```bash\ncodex --quiet \"Review the codebase structure and identify: \\\n 1. Circular dependencies \\\n 2. Missing error handling \\\n 3. Security risks \\\n Output as a prioritized list.\"\n```\n\n## Model Selection\n\n```bash\ncodex --model o3 # Best reasoning\ncodex --model o4-mini # Fast, cost-efficient\ncodex --model gpt-4.1 # General purpose\n```\n\nCodex defaults to the model associated with your subscription tier. Pro subscribers get access to o3 and full context.\n\n## Context Management\n\nCodex loads files from the current directory into context. To control what's loaded:\n\n```bash\n# Add specific files\ncodex \"With context from src/auth/ and src/api/, refactor the auth middleware\"\n\n# Ignore patterns (via .codexignore or .gitignore)\necho \"node_modules/\" >> .codexignore\necho \"*.log\" >> .codexignore\n```\n\n## Non-Interactive Mode\n\nFor CI pipelines and automation:\n\n```bash\n# Run single task, no prompts\ncodex --quiet --yolo \"Run lint and fix all errors\"\n\n# With specific model\ncodex --quiet --model o4-mini \"Generate TypeScript types from schema.sql\"\n```\n\n## Comparison: Codex vs Claude Code\n\n| Feature | Codex | Claude Code |\n|---------|-------|-------------|\n| Underlying LLM | OpenAI GPT/o-series | Anthropic Claude |\n| Tool approval UI | Inline diffs | Interactive prompt |\n| Sub-agents | No | Yes (.claude/agents/) |\n| Skills system | No | Yes (.claude/skills/) |\n| Memory | Session only | Persistent (MEMORY.md) |\n| MCP support | No | Yes |\n| Hooks | No | Yes (PostToolUse, Stop) |\n| Best for | Code-heavy focused tasks | Orchestration, multi-step |\n\n## Agency Config (Codex Desktop)\n\nCodex Desktop reads config from `~/.codex/` including agent definitions and skill references. Agents are loaded at startup and not hot-reloaded — restart required after config changes.\n","html":"

Overview

\n

OpenAI Codex CLI is a terminal-based AI coding assistant. It runs in a project directory and can read files, write code, run commands, and iterate based on test output.

\n

Installation

\n
npm install -g @openai/codex\n\n# Verify\ncodex --version\n
\n

Requires Node.js 18+.

\n

Authentication

\n

| Method | When |\n|--------|------|\n| ChatGPT login | Any ChatGPT subscription — browser OAuth redirect |\n| API key | No subscription, or programmatic access |

\n

Plans: Go ($8/mo), Plus ($20/mo), Pro ($100–$200/mo). Pro recommended for fast mode and high reasoning tasks.

\n

Commands Quick Reference

\n

| Codex | Claude Code Equivalent | Description |\n|-------|----------------------|-------------|\n| codex | claude | Launch agent in current directory |\n| codex --yolo | claude --dangerously-skip-permissions | Full access, no approval prompts |\n| codex --model o3 | --model claude-opus-4-6 | Specify model |\n| codex --quiet | n/a | Minimal output, automation-friendly |

\n

Sandbox Modes

\n

Codex runs in one of three modes, selected at startup:

\n

| Mode | Behavior | When to Use |\n|------|----------|-------------|\n| suggest (default) | Proposes changes, you approve each | Unfamiliar codebase, learning |\n| auto | Auto-applies low-risk changes, asks on risky ones | Day-to-day development |\n| full (--yolo) | No approvals, runs everything | Trusted projects, CI environments |

\n

Slash Commands

\n

Codex supports slash commands in the interactive session:

\n
/clear          — clear conversation context\n/history        — show conversation history\n/diff           — show uncommitted changes\n/files          — list files in context\n/model <name>   — switch model mid-session\n/quit           — exit\n
\n

Workflow Patterns

\n

Pattern 1: Plan First, Then Build

\n
codex \"Create a plan for adding Stripe subscriptions to this Next.js app. \\\n       List all files that need to change. Don't write code yet.\"\n\n# Review the plan\ncodex \"Execute the plan. Write all the code.\"\n
\n

Pattern 2: Test-Driven Development

\n
codex \"Write tests for the UserService class first. \\\n       Then implement the class to make them pass.\"\n
\n

Pattern 3: Iterative Bug Fix

\n
codex \"Run the test suite and fix any failing tests. \\\n       Keep running until all tests pass.\"\n
\n

Pattern 4: Architecture Review

\n
codex --quiet \"Review the codebase structure and identify: \\\n               1. Circular dependencies \\\n               2. Missing error handling \\\n               3. Security risks \\\n               Output as a prioritized list.\"\n
\n

Model Selection

\n
codex --model o3                   # Best reasoning\ncodex --model o4-mini              # Fast, cost-efficient\ncodex --model gpt-4.1              # General purpose\n
\n

Codex defaults to the model associated with your subscription tier. Pro subscribers get access to o3 and full context.

\n

Context Management

\n

Codex loads files from the current directory into context. To control what's loaded:

\n
# Add specific files\ncodex \"With context from src/auth/ and src/api/, refactor the auth middleware\"\n\n# Ignore patterns (via .codexignore or .gitignore)\necho \"node_modules/\" >> .codexignore\necho \"*.log\" >> .codexignore\n
\n

Non-Interactive Mode

\n

For CI pipelines and automation:

\n
# Run single task, no prompts\ncodex --quiet --yolo \"Run lint and fix all errors\"\n\n# With specific model\ncodex --quiet --model o4-mini \"Generate TypeScript types from schema.sql\"\n
\n

Comparison: Codex vs Claude Code

\n

| Feature | Codex | Claude Code |\n|---------|-------|-------------|\n| Underlying LLM | OpenAI GPT/o-series | Anthropic Claude |\n| Tool approval UI | Inline diffs | Interactive prompt |\n| Sub-agents | No | Yes (.claude/agents/) |\n| Skills system | No | Yes (.claude/skills/) |\n| Memory | Session only | Persistent (MEMORY.md) |\n| MCP support | No | Yes |\n| Hooks | No | Yes (PostToolUse, Stop) |\n| Best for | Code-heavy focused tasks | Orchestration, multi-step |

\n

Agency Config (Codex Desktop)

\n

Codex Desktop reads config from ~/.codex/ including agent definitions and skill references. Agents are loaded at startup and not hot-reloaded — restart required after config changes.

\n"}