A complete guide to the OpenAI Codex app and its macOS availability

OpenAI Codex has been a huge help for developers, acting as an AI coding partner that can write code, manage projects, and fix bugs. For a while, though, using it on a Mac was a bit of a disconnected experience. That all changed with the launch of a dedicated Codex app for macOS in February 2026.

What is the situation with Codex in 2026? This post will walk you through it. We'll cover what Codex actually is, explore the different ways you can use it on your Mac (including the new app and the command line), and clear up any confusion about its availability.

What is OpenAI Codex?

It's best to think of OpenAI Codex not as a single tool, but as a suite of AI agents built for software engineering. It does a lot more than just autocomplete your code. You can ask it to build a new feature, explain a confusing part of your codebase, debug tricky errors, and even help manage project tasks.

Under the hood, it’s powered by specialized models like GPT-5.2-Codex, which are fine-tuned for complex coding and agent-like behaviors. The cloud version you might use in your browser runs on codex-1, a custom version of OpenAI's o3 model.

To use it on a Mac, you have a few options, and we'll dig into each one:

• The new, dedicated Codex app for macOS.
• A lightweight, open-source Codex CLI for your terminal.
• A powerful, asynchronous cloud agent you can access at chatgpt.com.
• The "Work with Apps" feature inside the standard ChatGPT app for macOS.

How to use Codex on macOS

Developers have a few different ways to work with Codex, each tailored for a specific workflow. Whether you need a quick fix in the terminal or you're managing a long-running project, there’s an interface for that.

The dedicated OpenAI Codex app for macOS

Launched in February 2026, the new dedicated app is essentially a command center for your AI agents, designed from the ground up for macOS on Apple Silicon. It's built for handling serious development work.

Here are its core features:

• Parallel agents: You can run multiple AI agents at the same time in separate threads. Need one agent to work on a new feature while another refactors some old code? No problem. They won’t get in each other’s way or lose context.
• Isolated worktrees: Each agent works on its own isolated copy of your code. This is a big deal because it means you can review the diffs and merge changes without worrying about conflicts. It’s like each agent gets its own git branch.
• Skills & automations: You can extend Codex beyond just writing code by creating reusable "skills." For example, you could create a skill to generate images with GPT Image or one to deploy your project to Vercel. You can also set up "automations" to handle repetitive tasks, like triaging new issues every morning.

This app is the most powerful way to manage multiple AI agents on long-running, complex projects.

The Codex CLI for your terminal

If you live in the terminal, the Codex CLI is for you. It's a fast, open-source tool built in Rust that lets you collaborate with Codex in real-time. It’s available for macOS and Linux, with experimental support for Windows.

With the CLI, you can ask Codex to read and edit local files, run commands, and even perform code reviews right from your command line. It uses the heavy-duty GPT-5-Codex models by default, but you can switch to more efficient models like GPT-5.1-Codex-Mini to make your usage limits stretch up to four times longer.

The cloud-based agent in your browser

For big, complex tasks that don't need your constant attention, there's the cloud agent. You can access it at chatgpt.com/codex and delegate asynchronous jobs. You just give it a task, and the agent gets to work in a secure cloud environment that's preloaded with your GitHub repository.

This is perfect for things that might take a while. You can check in on its progress in real-time, and once it's done, you can review all the proposed changes and create a pull request with a single click.

The 'Work with Apps' feature in the ChatGPT app

This is a feature inside the standard ChatGPT for macOS app, not the new dedicated Codex app.

It uses the macOS Accessibility API to let ChatGPT see what's in your active application window. This means you can ask questions about your code without having to copy and paste anything. It’s handy for getting quick, contextual answers and works with popular apps like VS Code, Xcode, JetBrains IDEs, iTerm, Warp, Notion, and even Apple Notes.

OpenAI Codex availability on macOS

With all these different tools, it’s easy to get confused about what’s available and which one you should be using. Let's break it down.

Confirming the native macOS app

Previously, Codex was primarily accessed through a web interface. OpenAI officially launched a dedicated, feature-packed Codex app for macOS on February 2, 2026. This is now the main desktop experience for anyone doing serious, agent-based software development on a Mac. If you're on Windows or Linux, versions for those platforms are planned, and you can sign up to be notified when they're ready.

Codex availability across different methods

To make it crystal clear, here’s a quick summary of your options for using Codex on a Mac.

Method	Best For	How to Access
Codex App for macOS	Supervising multiple AI agents, long-running tasks, and using skills/automations.	Download the dedicated app.
Codex CLI	Quick, interactive coding assistance and scripting in the terminal.	Install via npm or brew.
Cloud Agent (Web)	Delegating complex, asynchronous tasks that can run in the background.	Visit chatgpt.com/codex in a browser.
'Work with Apps'	Getting quick, contextual answers about the code in your active editor.	Use the standard ChatGPT for macOS app.

Limitations of the current workflow

While Codex is incredibly powerful, the current setup is still a bit fragmented. A developer might use the dedicated app for a big feature, switch to the CLI for a quick refactor, and then jump over to the standard ChatGPT app to ask a quick question about an error message.

This constant context-switching between different interfaces can create friction and slow you down. And when you need answers that aren't in your codebase, like searching through your company's internal documentation, this fragmented approach can be a limitation. This highlights the benefit of having AI assistance integrated directly into existing workflows.

For instance, tools like eesel AI's Internal Chat integrate directly into places your team already works, like Slack or Microsoft Teams. This lets developers get instant answers from Confluence, Google Docs, or Notion without ever having to leave their chat window.

Getting started with Codex on macOS

Ready to jump in? Here’s how to get the different Codex tools set up on your Mac.

Installing the macOS app

1. First, make sure your Mac is running macOS 14+ and has Apple Silicon (an M1 chip or newer).
   

   Apple Silicone only :(

   Reddit

2. Head over to the [OpenAI desktop app page](https://openai.com/chatgpt/desktop/) and download the installer. 3. Open the `.dmg` file, drag the app to your Applications folder, and sign in with your ChatGPT account. 

Installing the Codex CLI

You can install the CLI with either npm or Homebrew. Just open your terminal and run one of the following commands.

Using npm:

npm install -g @openai/codex

Using Homebrew:

brew install codex

The first time you run the codex command, it will prompt you to sign in to your OpenAI account.

Setting up 'Work with Apps'

This one requires a few steps inside the standard ChatGPT for macOS app.

1. Open the ChatGPT app and go to Settings -> Work with Apps.
2. Toggle the "Enable Work with Apps" switch to on. Your Mac will ask you to grant Accessibility permissions, which is necessary for the feature to work.
3. If you're using VS Code, you'll also need to install the ChatGPT VS Code extension from the marketplace.

   

OpenAI Codex: Pricing and plans

Codex isn't a separate product you have to buy. Its functionality is included with ChatGPT subscriptions, and for a limited time, it's even available for Free and Go users to try out.

Here are the main subscription tiers that include Codex:

• Plus: $20/month. This is great for focused coding sessions. It comes with usage limits, but they're pretty generous (around 45-225 local messages every 5 hours).
• Pro: $200/month. Aimed at full-time developers, this plan gives you 6x higher usage limits than the Plus plan.
• Business: $30/user/month. This tier adds a secure workspace for your team, SAML SSO for easy sign-on, and faster processing for cloud tasks.

  

If you happen to burn through your usage limits on any of these plans, you have the option to purchase additional credits.

For developers who want to use the API directly, you'll pay per token. As an example, the gpt-5.2-codex model costs $1.75 per 1 million input tokens and $14.00 per 1 million output tokens.

For a more detailed look at the new Codex app in action, check out this introductory video from OpenAI. It provides a great first look at the interface and its core capabilities for managing AI agents on macOS.

A first look at the new OpenAI Codex app for macOS, showcasing its user interface and core capabilities.

The future of AI coding assistants

Thanks to the new dedicated app, using Codex on macOS is more powerful and straightforward than ever. However, the ecosystem is still spread across a few different tools, each designed for a specific job.

While Codex is a powerful AI teammate for writing code, other kinds of AI teammates are transforming how teams access knowledge and get work done. For development or support teams seeking to centralize knowledge access and manage repetitive inquiries, an integrated AI teammate can be a useful tool.

eesel AI is an AI teammate that integrates with tools like Zendesk and Slack. It learns from existing data sources such as help desk history and internal documentation to answer questions and assist with customer support, allowing teams to access information within their existing workflows.