OpenAI launched a standalone macOS app for Codex on Monday, opening its AI coding agent to all ChatGPT users temporarily while doubling rate limits for paid subscribers. The timing matters: AI coding tools are the current enterprise battleground, and OpenAI needs to defend against Anthropic's Claude and startup Cursor.
The app positions Codex as a "command center" for managing multiple AI agents in parallel across projects. Each agent runs in isolated cloud sandboxes and can handle tasks from bug fixes to PR generation, with OpenAI claiming "weeks of work in days." The interface organizes work by project threads, letting developers review changes as agents complete long-running tasks.
Codex runs on codex-1, OpenAI's optimized o3 model trained via reinforcement learning on real coding workflows. The system includes a Skills library for capabilities beyond code generation, like image creation, and emphasizes verifiable outputs with citations, logs, and tests. OpenAI positions this as addressing security concerns, though the company still flags it as a "research preview" requiring manual code review.
Access strategy is notable: ChatGPT Pro, Enterprise, Business, Plus, and Edu subscribers get Codex as standard, with the ability to buy additional credits. Free and Go tier users get temporary access with the Mac app launch, likely an acquisition play. Paid tiers get 2x rate limits temporarily.
CEO Sam Altman called it "the most loved internal product we've ever had" during a Friday briefing. The enthusiasm is strategic: OpenAI claims 1 million developers used Codex last month since its October general availability, but Anthropic and Cursor have momentum with their own agentic coding tools.
The Mac-only launch is a constraint worth noting. Cross-platform availability will determine whether this is a developer tool or a platform play. The doubled limits and free access won't last forever, so enterprise teams evaluating AI coding assistants should test now while the economics are temporarily favorable.
What to watch: How Codex pricing compares to GitHub Copilot Enterprise and Anthropic's offerings when promotional rates end, and whether the multi-agent approach delivers on the "weeks to days" claim in production environments.