OpenAI Codex App: Complete Guide to the New Multi-Agent Coding Desktop

OpenAI launched the Codex App for macOS - a desktop application for multi-agent coding. Everything about installation, skills, automations, and how it compares to Claude Code.

OpenAI Codex App: Complete Guide to the New Multi-Agent Coding Desktop

OpenAI Codex App: Complete Guide to the New Multi-Agent Coding Desktop

Codex App, OpenAI's multi-agent coding desktop application, launched as a new macOS app yesterday (February 2, 2026). Here's everything you need to know — from installation to advanced automations.


What is the OpenAI Codex App?

Codex App the Codex App is OpenAI's new macOS desktop application designed specifically for working with multiple AI coding agents. Unlike the existing Codex CLI and IDE extensions, the app provides a dedicated interface to manage multiple agents in parallel, supervise long-running tasks, and orchestrate complex projects.

The Codex App is OpenAI's desktop coding assistant that lets developers build software with AI agents. As Sam Altman puts it: "You can use this from a clean sheet of paper, brand new, to make a really quite sophisticated piece of software in a few hours. As fast as I can type in new ideas, that is the limit of what can get built."

Why a Desktop App?

The way developers work with AI agents has fundamentally changed. Models are now capable of handling complex, long-running tasks end to end. Developers are now orchestrating multiple agents across projects: delegating work, running tasks in parallel, and trusting agents to take on substantial projects that can span hours, days, or weeks.

The core challenge has shifted from what agents can do to how people can direct, supervise, and collaborate with them at scale. Existing IDEs and terminal-based tools aren't built to support this way of working.


Availability and Pricing

PlanCodex AccessSpecial Notes
ChatGPT Free/Go✅ AvailableTemporarily included for limited time
Plus✅ IncludedRate limits doubled
Pro✅ IncludedRate limits doubled
Business✅ IncludedRate limits doubled
Enterprise✅ IncludedRate limits doubled
Edu✅ IncludedRate limits doubled

Download: The app is available at openai.com/codex.

Platforms: Currently macOS only. A Windows version is in development.


Installation and Getting Started

1. Download the App

  1. Codex App Go to openai.com/codex
  2. Download the macOS app
  3. Install by dragging to the Applications folder
  4. Launch and sign in with your ChatGPT account

2. Seamless Integration

A major advantage: The app automatically picks up your session history and configuration from the Codex App CLI and IDE extension. You can immediately start working with your existing projects.

3. Authentication

You can use Codex in two ways:

  • ChatGPT Login (recommended): Uses your Plus, Pro, Team, Edu, or Enterprise subscription
  • API Key: For developers who need programmatic access (requires additional setup)

Key Features

1. Multi-Agent Command Center

The Codex App is a command center for agents. Codex App You can:

  • Run multiple agents in parallel — each in a separate thread, organized by project
  • Switch between tasks without losing context
  • Review changes in the thread, comment on diffs, and open directly in your editor

2. Worktrees for Conflict-Free Work

Built-in worktree support means multiple agents can work on the same repository without conflicts. Each agent works on an isolated copy of your code.

Workflow example:

Agent 1: Auth module refactoring → worktree/auth-refactor
Agent 2: New API endpoints → worktree/api-endpoints  
Agent 3: Bug fixes → worktree/bugfixes

→ All working in parallel, no Git chaos

You can:

  • Check out changes locally
  • Or let the agent continue making progress without touching your local Git state

3. Skills: Extending OpenAI Codex App

Skills are folders containing instructions, resources, and scripts that Codex can use to reliably complete specific tasks. They enable Codex to go beyond pure code generation.

Pre-installed Skills (Selection)

SkillFunction
Figma IntegrationFetch design context and assets from Figma, translate to production-ready UI code
Linear IntegrationBug triage, release tracking, team workload management
Cloud DeploymentDeploy directly to Cloudflare, Netlify, Render, or Vercel
Image GenerationCreate images with GPT Image for websites, UI mockups, game assets
OpenAI DocsAccess up-to-date OpenAI API documentation while developing
Document CreationCreate and edit PDF, spreadsheet, and DOCX files

Creating Custom Skills

Skills are automatically recognized everywhere you use Codex — in the app, CLI, or IDE. You can also check skills into your repository to share them with your entire team.

# Install a skill (example)
$skill-installer gh-address-comments

# Skill from experimental folder
$skill-installer install https://github.com/openai/skills/tree/main/skills/.experimental/create-plan

Skills Repository: github.com/openai/skills

4. Automations: Background Work

With Automations, you can let Codex work in the background on an automatic schedule. When an automation finishes, the results land in a review queue.

Typical automation use cases at OpenAI:

  • Daily issue triage
  • Finding and summarizing CI failures
  • Generating daily release briefs
  • Running bug checks
  • Alert monitoring

Setting up an Automation:

  1. Define instructions + optional skills
  2. Set a schedule
  3. Let Codex work
  4. Review results when done

5. Personalities: Your Working Style

Developers have different preferences. Codex now offers two personality modes:

ModeDescription
PragmaticShort, execution-focused, minimal explanations
EmpatheticMore communicative, explanatory, friendly tone

Switch with the /personality command — works in app, CLI, and IDE.


Security Model

The Codex App uses native, configurable system-level sandboxing (Open Source: github.com/openai/codex).

Default Restrictions

  • Agents can only edit files in the current project folder/branch
  • Only cached web search allowed
  • Commands with elevated permissions (e.g., network access) require approval

Configurable Permission Levels

LevelDescription
UntrustedMaximum restrictions
On failureAsk when something fails
On requestAsk when elevated permissions needed
NeverApp may never ask for elevated permissions

You can define rules for your project or team that allow certain commands to automatically run with elevated permissions.


Practical Example: A Racing Game from One Prompt

OpenAI demonstrated what's possible: A complete racing game — with different racers, eight maps, and items — was created with just one initial prompt.

The agent used:

  • Image Generation Skill (powered by GPT Image)
  • Web Game Development Skill

Statistics:

  • Over 7 million tokens processed
  • Agent took on roles of designer, game developer, and QA tester
  • Validated work by actually playing the game

Where OpenAI Codex App Works Everywhere

The Codex App is part of an ecosystem — you can switch between different surfaces, and Codex maintains context:

SurfaceDescription
Codex AppCommand center for multi-agent workflows
Codex CLITerminal-based work
IDE ExtensionVS Code, Cursor, Windsurf, JetBrains IDEs
Codex WebCloud-based at chatgpt.com/codex

All connected through your ChatGPT account.


OpenAI Codex App vs Claude Code: Key Differences

AspectCodex AppClaude Code
ModelGPT-5.2-CodexClaude Opus 4.5
InterfaceDesktop App, CLI, IDE, WebDesktop App (Code Cowork), CLI
Multi-Agent✅ Native design✅ Via worktrees
SkillsOpen-source skills ecosystemIntegrated skills system
Automations✅ Built-in❌ Not native
PricingIncluded in ChatGPT subscriptionSeparate billing
Open Source CLI✅ github.com/openai/codex✅ Via npm

What's Coming Next?

OpenAI has outlined the roadmap:

  1. Windows version — In development
  2. Better model capabilities — Continuous improvements
  3. Faster inference — Performance optimizations
  4. Cloud-based triggers — Automations run continuously, not just when your computer is open
  5. Improved multi-agent workflows — Based on real-world feedback

Usage numbers (as of February 2026):

  • Over 1 million developers using Codex
  • Usage has doubled since GPT-5.2-Codex launch
  • Enterprise customers: Cisco, Ramp, Virgin Atlantic, Vanta, Duolingo, Gap

OpenAI Codex App: Conclusion: Is It Worth Switching?

The Codex App represents a fundamental shift: from the pair-programming metaphor (one agent, one task) to orchestrating agent teams across the entire software lifecycle.

Who is the Codex App ideal for?

  • Teams wanting to coordinate multiple agents in parallel
  • Developers with long-term refactoring or migration projects
  • Companies wanting automated code reviews and CI workflows

Who should wait?

  • Developers who only occasionally need AI assistance
  • Teams already deeply invested in Claude Code
  • Windows users (until the Windows version releases)

The temporary availability for Free/Go users and doubled rate limits make now the perfect time to try the Codex App.


This article was published on February 3, 2026, one day after the Codex App launch. Information is based on the official OpenAI announcement, technical documentation, and coverage from TechCrunch, Ars Technica, ZDNet, and other sources.

Share article

Share: