OpenAI ships Codex macOS app: multi-agent command center with git worktrees and skills
OpenAI introduced the macOS-only Codex app as a "command center" to run multiple coding agents in parallel, isolate work via git worktrees, and extend workflows with a new Skills system—plus a limited-time inclusion with ChatGPT Free/Go and doubled rate limits for paid plans ([OpenAI blog](https://openai.com/index/introducing-the-codex-app/?_bhlid=b040462c226c34eb9531cc536689e69b976397a7)[^1]). Developer docs confirm Apple Silicon support today, a Windows/Linux waitlist, and that API-key sign-in may limit features like cloud threads ([Codex app docs](https://developers.openai.com/codex/app/)[^2]). Reporting adds competitive context against Anthropic’s Code Cowork/Claude Code and notes model guidance (use GPT‑5.2‑Codex for coding) and multi-agent monitoring aimed at centralizing team workflows ([Fortune](https://fortune.com/2026/02/02/openai-launches-codex-app-to-bring-coding-models-to-more-users-openclaw-ai-agents/)[^3]). [^1]: Adds: official product details on multi-agent orchestration, git worktrees, Skills, and rate limit changes. [^2]: Adds: confirms macOS-only (Apple Silicon), Windows/Linux waitlist, and API-key limitations for cloud threads. [^3]: Adds: market context vs Anthropic, enterprise adoption, model recommendations, and multi-agent monitoring pitch.