OpenAI has officially entered a new competitive arena with the launch of a dedicated desktop application for its Codex AI coding assistant. The new Codex app for macOS is designed as a "command center" for managing multiple AI coding agents simultaneously, allowing developers to run tasks in parallel and collaborate with AI on complex, long-running projects.

The app aims to change how software is built by shifting from simple code suggestions to orchestrating teams of AI agents. It provides a central interface where developers can create separate project threads, review the code changes made by each AI agent, add comments, and merge approved work into the main codebase. A key feature is support for Git worktrees, which lets each AI agent work on an isolated copy of the code to prevent conflicts.

To broaden its capabilities, Codex now supports "Skills"—pre-configured bundles of instructions and scripts that allow it to perform tasks beyond just writing code. This includes generating images, building web games, and synthesizing information. For a limited time, OpenAI is including Codex access for users on its free ChatGPT tiers and is doubling the usage limits for paying customers on Plus, Pro, Business, and Enterprise plans.

The launch comes as competition intensifies in the AI assistant space, particularly from Anthropic's Claude Code and Claude Cowork apps, which have gained rapid adoption. OpenAI CEO Sam Altman emphasized that the app combines the raw power of their latest GPT-5.2-Codex model with a more intuitive interface, claiming it allows a developer to build sophisticated software "from a clean sheet of paper" in just a few hours.

This move positions OpenAI to directly compete for developer attention on the desktop, where rivals have been building momentum. The strategy is to make advanced, agentic AI development feel less like juggling browser tabs and more like directing a coordinated team.