Kenhuang Academy
  • English

      • Complete Codex Tutorial

      Welcome to the Complete Codex Tutorial. This tutorial is based on OpenAI's official documentation (developers.openai.com/codex) and the open-source implementation (openai/codex). It uses the Codex Desktop App as the main thread while also covering Codex CLI, IDE extensions, and Codex Cloud, helping you move from "installed and running" to deeper customization, team rollout, and automation.

      Codex Desktop App main interface

      The Codex Desktop App new-thread screen: the left sidebar contains threads, search, plugins, automations, tasks, projects, and cloud entry points. The center area contains the prompt box plus permission, model, and reasoning-effort selectors.

      Who This Tutorial Is For

      • Developers who want to drive AI coding through a visual desktop app, not only terminal commands.
      • Existing Codex CLI users who want a systematic view of the App, Cloud, Computer Use, Plugins, and newer capabilities.
      • Technical leads who need to roll Codex out for a team or enterprise using AGENTS.md, Skills, permissions, rules, and hooks.

      Tutorial Map

      1. Getting Started

      ChapterWhat You Will Learn
      Introduction to CodexWhat Codex is, the product matrix (App / CLI / IDE / Cloud), and account basics
      Core ConceptsThe agent loop, instruction chain, progressive disclosure, three permission gates, and five customization layers
      Installation & UsageDownload the desktop app, install the CLI, and sign in
      Desktop AppFirst launch, adding projects, Local/Cloud, handoff, and parallel threads
      App InterfaceSidebar, thread area, diff review, top bar, bottom status bar, and selectors
      Quick StartComplete one verifiable task in the App and one in the CLI

      2. Configuration & Customization

      ChapterWhat You Will Learn
      Configuration & Customizationconfig.toml layering, profiles, and CODEX_HOME
      PermissionsSandbox modes, approval policies, and Auto / Read-only / Full access presets
      AGENTS.mdDiscovering, merging, and rolling out persistent project instructions
      Agent SkillsSKILL.md, progressive disclosure, plugins, and the plugin marketplace
      Slash Commands/status, /model, /approvals, /review, /goal, and more
      Rules & HooksStarlark command rules and lifecycle hooks

      3. Advanced Capabilities

      ChapterWhat You Will Learn
      Computer UseSeeing the screen, operating GUIs, the in-app browser, Appshots, and Goal mode
      Model SelectionGPT-5.5 / 5.4 / 5.4-mini / Codex Spark and reasoning effort
      SubagentsExploration, review, and MCP-specialized subagent patterns
      MCP IntegrationConnecting Codex to external systems through MCP
      Integrations & ConnectionsPlugin marketplace, Apps/connectors, Figma, Linear, Slack, and GitHub
      Workflows & WorktreesParallel threads, worktree isolation, and plan → implement → review workflows
      Automation & CI/CDApp Automations, non-interactive mode, and GitHub Actions
      Prompting Best PracticesHow to write high-hit-rate instructions
      Advanced TipsContext management, checkpoints, resume/fork, @ mentions, and images

      4. Surfaces, Security & Practice

      ChapterWhat You Will Learn
      IDE ExtensionUsing Codex in VS Code / Cursor
      Codex Cloudchatgpt.com/codex, @codex, PRs, and mobile use
      Security & Enterprise ManagementSandboxing, requirements.toml, managed configuration, and auditability
      Practical ExamplesRefactoring, automated review, and Issue → PR workflows
      CheatsheetCommands, slash commands, config, and models on one page
      Command Line Tool (CLI)CLI config, sandboxing, models, Skills, prompts, and practical examples
      ResourcesOfficial docs and community references

      Suggested Learning Path

      StageGoalRecommended Reading
      Day 1Install the desktop app and finish the first small taskIntroduction → Installation → Desktop App → Quick Start
      Days 2-3Understand the interface and how the agent worksApp Interface → Core Concepts → Permissions
      Week 1Roll out conventions in a team repoAGENTS.md → Skills → Slash Commands → MCP
      AdvancedUse the newer capabilities wellComputer Use → Integrations → Workflows → Automation
      OngoingScale safely and govern usageSecurity & Enterprise → Rules & Hooks → CLI → Practical Examples

      Prerequisites

      • Familiarity with terminal and Git basics. Many App features are Git-based.
      • At least one programming language.
      • A ChatGPT Plus / Pro / Business / Edu / Enterprise plan, or OpenAI API credits.

      Version note: Codex changes quickly. Model names, feature flags, and UI details may change. Treat the official documentation as the source of truth. This tutorial is based on public documentation available in 2026.

      评论