OpenAI launches Codex as a new Mac app for developers Apfelpatient

OpenAI is expanding its macOS offerings with a new application: Codex is getting its own Mac app. The app is clearly aimed at developers who work with agents and want to streamline their workflow between AI, IDE, and terminal. OpenAI describes Codex as the best way to develop using agents.

Until now, Codex was primarily available through CLI and IDE extensions. With the new macOS app, OpenAI is taking a further step, delivering a standalone platform designed not just as a supplement, but as a central work environment. Alongside ChatGPT and Atlas, Codex is OpenAI's third macOS app and is clearly focused on productive work with code and agents.

Codex is getting its own Mac app

With the release of the Codex app for macOS, OpenAI has created a dedicated space for agent-based programming. The app is designed to integrate seamlessly into existing development environments. Switching between the Codex app, IDEs, and the terminal is effortless.

Codex is not intended as a simple frontend, but as an independent platform on which complex development tasks can be organized and processed in parallel.

Multitasking with agents and a clear structure

A key feature of the Codex app is its dedicated space for multitasking. Agents run in separate threads, organized by project. This ensures that the context of each task is maintained, even when multiple agents are active simultaneously.

You can switch between these threads at any time without losing any information. Changes made by an agent can be reviewed directly in the respective thread. Differences can be commented on and, if necessary, opened in a separate editor for manual revision or fine-tuning.

Collaboration via Worktrees without conflicts

The Codex app includes integrated support for Git worktrees. Multiple agents can work on the same repository in parallel without blocking each other or causing conflicts.

Each agent works on an isolated copy of the code. This allows different approaches to be tested simultaneously without constantly having to monitor their impact on the main codebase. While an agent is running in the background, changes can be reviewed locally, or the agent can continue running without affecting the local Git state.

Seamless takeover of existing projects

A key advantage for practical use is the ability to import existing setups. The Codex app adopts session history and configuration from the Codex CLI and the IDE extension. This allows the app to be used immediately with existing projects without reconfiguration or duplicate setup.

Background automation

Additionally, the Codex app supports automations. These allow Codex to run in the background according to a set schedule. An automation combines specific instructions with optional capabilities and is executed automatically.

Once an automation process is complete, the results are placed in a review queue. From there, work can be resumed and continued if needed, without having to restart the entire process.

Availability and extended access at launch

To coincide with the launch of the Codex app, OpenAI is expanding access to its features. The app is now available to users of ChatGPT Free and ChatGPT Go. Additionally, rate limits for Plus, Pro, Business, Enterprise, and Edu users are being temporarily doubled. This will allow more users to test Codex more extensively and integrate it into existing workflows.

The Codex app for macOS can be downloaded directly from the official OpenAI website. A detailed blog post is also available, providing further details and background information about the app.

OpenAI expands the macOS ecosystem with Codex

With the Codex app for macOS, OpenAI is strategically expanding its ecosystem and providing a specialized development environment for agent-based programming. The focus on structured threads, parallel agents, isolated worktrees, and time-based automation clearly indicates the direction of development. Codex is intended not only to provide support but also to serve as a central hub for modern software development on the Mac. (Image: OpenAI)