Skip to content

swhan0329/codex-101

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

46 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Codex 101

Codex 101 — OpenAI Codex Complete Guide

The definitive bilingual (🇰🇷 KR / 🇺🇸 EN) guide to OpenAI Codex
CLI · Desktop App (macOS/Windows) · IDE Extension · Web Dashboard

Last updated: April 19, 2026
The live site displays the viewer's current date automatically and reflects the latest Codex model guidance reviewed from official docs.

🌐 Live Site · 🇰🇷 한국어


📖 About

Codex 101 is a comprehensive guide to OpenAI's coding agent Codex, organized for both first-time users and advanced practitioners.

Based on the official OpenAI documentation and OpenAI Docs MCP lookups, the guide is manually reviewed and verified before publication.

Quick Paths By Audience

  • First-time users: Start with sections 04-06 (setup, sign-in, first run), then 10 (approval/sandbox basics), and 14 (OpenAI Docs MCP).
  • Professional users: Start with sections 12-14 (AGENTS.md, config.toml, MCP), then 15-17 (session strategy, automation, prompting contracts).

Daily MCP Verification Snapshot (2026-04-19)

  • Re-checking codex/models still shows the same four recommended lanes: gpt-5.4, gpt-5.4-mini, gpt-5.3-codex, and gpt-5.3-codex-spark. The important current change is that the Codex models page now effectively leaves gpt-5.2 as the main alternative model, while gpt-5.3-codex is explicitly described as the coding capability that also powers GPT-5.4.
  • Reading Introducing GPT-5.4 together with Introducing GPT-5.4 mini and nano sharpens the model story further: gpt-5.4 is the flagship professional model for Codex, gpt-5.4-mini is the fast local/subagent choice, and nano still does not appear on the Codex models page as a recommended or alternative Codex surface model.
  • Quickstart and Pricing remain the most important moving pieces. Quickstart still says every ChatGPT plan includes Codex, while Pricing currently distinguishes Pro $100 (10x through May 31, 2026) from Pro $200 (20x by default plus a temporary 25x five-hour boost through the same date). Business and Enterprise can still add Codex-only pay-as-you-go seats, and eligible Business workspaces can earn up to $500 in credits while token-based billing expands.
  • The setup story is clearer than older revisions: Quickstart treats the app as the easiest starting path, the IDE surface explicitly covers VS Code, Cursor, Windsurf, and JetBrains, and API-key sign-in is still allowed but can limit features such as cloud threads.
  • Windows guidance is now split more cleanly between the general Windows page and the dedicated Windows app page. The default recommendation is the native app, installable from Microsoft Store or winget, with elevated sandbox preferred over unelevated, a private desktop on by default, and WSL positioned as the right option when your workflow already lives in Linux or the IDE agent path needs it.
  • The Codex app docs are broader than a simple desktop shell. The current app, app/features, and app/automations pages emphasize built-in worktrees, Automations, thread-vs-standalone automation choices, Git tools, integrated terminal, voice dictation, pop-out windows, IDE sync, image input, chats, artifact preview, cached web search by default for local tasks, notifications, and sleep prevention. They also clarify that project-scoped automations need the app running and the project available on disk, and that unattended automations try approval_policy = "never" when policy allows it.
  • Plugins are now documented as a real packaging layer, not just a passing mention. The official Plugins page clearly defines plugins as bundles of skills, app integrations, and MCP servers, and documents both the app plugin directory and the CLI /plugins flow.
  • config.toml is still the area most likely to drift. The current reference gives more weight to review_model, top-level web_search, tools.web_search, personality, service_tier, default_permissions, tools.view_image, windows.sandbox, windows.sandbox_private_desktop, model_instructions_file, granular approvals, app permissions, feature flags, and named permission profiles.
  • A wider sweep across developers.openai.com/codex and openai.com/index still did not justify a new top-level chapter today, but it did confirm that the guide should keep App-first onboarding, Docs MCP grounding, and model-selection guidance at the top while using Codex for OSS, the AI-native engineering team guide, and launch posts as supporting references. The OSS page is now especially worth noting because it highlights six months of ChatGPT Pro with Codex plus conditional Codex Security access for eligible maintainers.

Audience Quick Use

  • First-time users (10 minutes): install/sign in (04-06) → keep default sandbox/approval (10) → run one small edit task with Git checkpoint.
  • Professional users (team rollout): lock AGENTS.md + .codex/config.toml (12-14) → standardize codex exec/review flow (15-16) → adopt section 17 prompt contract as team baseline.

What's Covered

Section Topic
Start Here Fast paths for first-time users and professional workflows
01–03 Ecosystem overview, product suite, supported models
04–05 System requirements & pricing, installation & auth
06–09 CLI, App, IDE Extension, Web usage
10–14 Approval modes, slash commands, AGENTS.md, config.toml, MCP
15–16 Session management, CI/CD automation
17 Prompting GPT-5.4 overview with embedded contracts, workflows, and playbooks
18–19 Advanced techniques, Codex vs Cursor vs Claude Code
20–21 FAQ, references

🌍 Languages

Currently supports Korean 🇰🇷 and English 🇺🇸.
Toggle languages in real-time via the 🌐 button on the top right of the page.

We welcome translation contributions! Add a new language block in i18n.js.


⚠️ Disclaimer

The initial draft of this guide was created using AI based on the official OpenAI documentation and community resources, then manually reviewed and verified before publishing.
Despite our review, some typos, mistranslations, or inaccuracies may still remain.

If you find any issues or have questions, please open an Issue or submit a PR!
Codex is evolving rapidly — always check the official docs for the latest information.


🤝 Contributing

Open source contributions are welcome! 🎉

Ways to Contribute

  • 🐛 Bug fixes — Incorrect info, typos, mistranslations
  • 🌐 Translations — Add new language support (add to i18n.js)
  • 💡 Tips & workflows — Share useful Codex tips
  • 📸 Screenshots — Updated UI screenshots
  • 📝 Content — Expand or improve existing sections

How to Contribute

# 1. Fork & Clone
git clone https://github.com/<your-username>/codex-101.git
cd codex-101

# 2. Open in browser to verify changes
open index.html

# 3. Submit a PR
git checkout -b fix/my-improvement
git commit -m "Fix: correct outdated information"
git push origin fix/my-improvement

🏗️ Project Structure

codex-101/
├── index.html      # Main page (chapter-map overview + 21 sections, including Prompting GPT-5.4)
├── style.css       # Styles (dark/light theme)
├── app.js          # Interactions (theme, language, section navigation)
├── i18n.js         # Translations (KR/EN)
└── images/         # Screenshots

📝 License

This project is open source and available under the MIT License.


Made with ❤️ by @swhan0329
Built with the help of AI — PRs welcome!

About

OpenAI Codex Complete Guide 101 — Bilingual (KR/EN) GitHub Pages site

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors