OpenAI launches the Codex app for macOS and opens free access to any user (for a limited time)

OpenAI has introduced the Codex app for macOS today, “a new and powerful interface designed to effortlessly manage multiple agents at once, run tasks in parallel, and collaborate with agents on long-running assignments”, as the company stated in an announcement. This tool seeks to transform the way software is developed, offering support for one-off edits as well as coordination of teams of agents throughout the entire lifecycle of a project.

At its launch in April 2025, OpenAI presented Codex as a cloud-based software engineering agent that can work on many tasks in parallel. Thus, Codex can write functions, answer questions regarding your codebase, and each task is performed within its own sandbox environment in the cloud.

Now, the models are capable of handling complex projects from start to finish, while developers supervise and coordinate multiple agents on different assignments. With this advancement, the main challenge is no longer what agents can accomplish but rather how humans can direct, supervise, and collaborate with them at scale. Yet, according to OpenAI, terminal-based tools are not equipped for this type of work, which underscores the need for a new solution: the Codex app.

We’re excited to launch the Codex app, a command center for building with agents.

It gives you a focused space to manage multiple agents at once, run work in parallel, and collaborate with agents over long-running tasks.https://t.co/ldE9k0uL5z pic.twitter.com/pH3K6d9D3q

— OpenAI Developers (@OpenAIDevs) February 2, 2026

Managing multiple Agents in parallel

One of the main innovations of the application is its capability for multitasking with agents. Each agent operates in a separate thread and is organized by projects, enabling task-switching without losing context. In addition, developers can review each agent’s changes, comment on differences, or edit them directly within their development environment.

The app also provides integrated support for work trees, allowing several agents to collaborate within a single repository without conflicts. Each agent works in an isolated copy of the code, making it easier to explore various approaches without affecting the primary codebase. Additionally, the app collects session histories as well as CLI (command-line interface, where you enter text instructions) and IDE (integrated development environment, where you write code) extension settings, facilitating integration with existing projects.

Beyond code generation

Since its launch, Codex has evolved from being an agent that merely writes code to an assistant capable of executing tasks on your computer. So-called “skills” allow the agent to connect tools, run workflows, and autonomously complete tasks. For instance, OpenAI has demonstrated how Codex can create a complete racing game, taking on the roles of designer, developer, and quality assurance tester, managing millions of tokens with just a single initial instruction.

As the company explains,

“Skills group instructions, resources, and scripts so that Codex can reliably connect to tools, run workflows, and complete tasks based on your team’s preferences. The Codex app includes a dedicated interface to create and manage skills. You can explicitly request Codex to use specific skills or allow it to do so automatically, depending on the task at hand.

We asked Codex to create a racing game with different racers, eight maps, and even power-ups that players could use with the space bar. By leveraging an image generation skill and a web game development skill, Codex independently created the game, utilizing over 7 million tokens with a single initial prompt. The agent assumed the roles of designer, developer, and QA tester to validate its work through actual gameplay.”

The application includes a skills library that ranges from implementing Figma designs to project management in Linear, image generation with GPT Image, or deploying web applications on platforms like Vercel or Netlify. These skills can be utilized in the app, the CLI, the IDE extension, or shared in a repository for an entire team to access.

Automations and personalization

Codex allows you to configure automations that execute tasks according to a schedule, such as generating reports, error verification, or sorting issues. Furthermore, developers may choose from two agent personalities: either a concise and pragmatic style or a more conversational and empathetic one, enabling interaction to be tailored to their preferred working style.

With this new application, OpenAI aims to bridge the gap between what advanced AI models are capable of and how readily developers can apply them, delivering a powerful and versatile agent that not only writes code but also drives all kinds of technical knowledge work.

Availability and plans

The new application is available today for macOS and integrates with subscriptions such as ChatGPT Plus, Pro, Business, Enterprise, and Edu. Moreover, OpenAI has announced that, for a limited time, Codex will have universal availability for all users, regardless of their ChatGPT plan (although access to the new features exclusive to the macOS app is not included).

Therefore, ChatGPT Free and Go users will also have access to Codex, which means that anyone with a free account can test its capabilities, even without the macOS app. In this way, users of Windows, Linux, or any other operating system may work with Codex agents without the need for a desktop app. During this promotional period, rate limits have been doubled for all paying users, improving the tool’s overall experience and performance.

In any case, OpenAI plans to expand the application to Windows and will continually enhance multi-agent workflows and cloud automation capabilities.

Image: OpenAI