OpenAI has released a new desktop version of Codex for macOS, giving developers a dedicated space to manage AI-powered coding work and keep track of longer software projects.
The app was announced on February 2 and will be available to macOS users starting February 3. OpenAI said access is included with certain ChatGPT subscriptions, with the option to purchase extra credits for higher usage.
Rather than working inside traditional development tools, Codex is positioned as a central workspace where developers can assign tasks to multiple AI agents, review code updates, and switch between ongoing jobs without losing progress. The app supports isolated worktrees, allowing several agents to work on the same codebase at the same time without clashing.
A Central Workspace for Coordinating AI Agents
OpenAI said developers are increasingly using AI for work that runs across days or even weeks. As a result, the main challenge is no longer what AI models can handle, but how people oversee and coordinate several agents working in parallel.
Alongside the macOS launch, OpenAI has expanded access to Codex across its subscription plans. For a limited time, Codex will also be available to ChatGPT Free and Go users. Usage limits have been increased for Plus, Pro, Business, Enterprise, and Edu subscribers across the desktop app, command-line tools, IDE extensions, and cloud usage.
The company said the Codex app supports a broader shift in how software is built. Developers can use it for quick code changes with a single agent or to supervise multiple agents working together across design, development, testing, and maintenance. OpenAI noted that most existing IDEs and terminal-based tools were not designed for this kind of workflow.
Codex also introduces “skills,” which bundle instructions and scripts so the system can interact more reliably with external tools. With skills enabled, Codex can help with tasks beyond writing code, including research, documentation, image creation, deployment, and basic project coordination. Developers can choose skills themselves or allow Codex to select them based on the task.
OpenAI said it already uses Codex internally for everyday work such as drafting documentation, managing issue backlogs, running evaluations, and preparing reports. The app includes a shared skills library and allows teams to reuse custom skills across projects.
Another feature, called Automations, lets developers schedule background tasks to run on a set timetable. Once a task finishes, the results are sent to a review queue. OpenAI said this is used internally for routine jobs like daily issue reviews and tracking build failures.
Developers can also switch between two interaction styles — a short, task-focused mode and a more conversational one — using a single command across the app, command-line tools, and IDE extensions.
On the security side, OpenAI said the Codex app uses system-level sandboxing similar to its existing developer tools. By default, agents are restricted to specific folders or branches and must ask for approval before taking higher-risk actions such as network access. Teams can adjust these rules to suit their needs.
OpenAI said it plans to release a Windows version of the Codex app in the future, expand automation features using cloud-based triggers, and continue improving how multiple AI agents work together on large software projects.
Also Read: OpenAI Introduces Age Prediction on ChatGPT to Strengthen Teen Safeguards
