OpenAI Introduces Codex App for macOS: What’s New and What It Can Do

OpenAI has announced the Codex app for macOS, aiming to help developers handle several AI agents together, run work in parallel, and manage long-running tasks. Codex access is bundled with most ChatGPT plans for now, and rate limits on paid tiers are being increased to support heavier development workloads across platforms.

During this introductory period, Codex is included with ChatGPT Free and Go plans, alongside Plus, Pro, Business, Enterprise and Edu subscriptions. OpenAI is doubling Codex rate limits for all paid plans, and these relaxed limits apply wherever Codex runs, including the desktop app, command line, IDE integrations and cloud environments.

How the Codex app reshapes software development

The Codex app is designed as a command centre for AI agents, changing how software is created and maintained. Developers can pair with a single agent for precise edits or oversee coordinated teams of agents that help with design, development, shipping and long-term upkeep across entire projects, including tasks that run for days or weeks.

OpenAI says Codex no longer just writes code snippets; it now executes work on the computer using code. Supported tasks extend beyond pure programming to information gathering, synthesis, structured problem-solving and writing. This approach reflects growing reliance on agents for multi-step workflows that demand context, planning and steady progress across longer timeframes.

Multi-agent workflows inside the Codex app

The Codex app focuses on multi-tasking with several agents at once. Each agent operates in its own thread, grouped under projects, so developers can jump between tasks without losing context. Within these threads, users can inspect proposed changes, comment on diffs, and open related files directly in local editors for manual review or adjustments.

Worktrees are built into the Codex app so multiple agents can safely work on one repository without clashing. Every agent edits an isolated copy of the codebase, allowing different approaches to be tested in parallel. Developers can choose to check out an agent’s changes locally or allow further automated progress while keeping the local git state untouched.

Codex app skills for design, documents and deployments

Skills extend what Codex can do beyond code generation, and the Codex app ships with a library focused on real workflows. These skills cover design translation, project management, deployments, image generation, API usage and office documents. The aim is to let agents move between planning, implementation and presentation work within the same environment.

New skills authored inside the Codex app become available wherever Codex operates: inside the desktop interface, from the Codex CLI or through the IDE extension. Teams can also commit shared skills into repositories, letting every contributor reuse the same tools and workflows when collaborating on larger software projects.

Automations and integrations in the Codex app

Automations in the Codex app let agents run on schedules without constant supervision. Each Automation combines a set of written instructions with optional skills, then runs at chosen intervals. When an Automation finishes, its output appears in a review queue, where developers can inspect the work, make edits or continue the process interactively.

The app reuses session history and configurations already set up via the Codex CLI and IDE extension, so existing users can adopt the desktop tool without reconfiguring projects. By tying together project threads, skills, Automations and higher rate limits, Codex is positioned as a central workspace for developers who rely on agents for sustained software work.

Via