OpenAI has introduced a new macOS application designed to integrate its Codex tool, aiming to facilitate the growing adoption of agentic coding approaches. The application marks a significant step towards aligning with the evolving landscape of software development.
Agentic coding represents a paradigm shift in software development, where AI agents autonomously tackle coding tasks, a concept showcased by apps like Claude Code and Cowork. OpenAI has been progressively expanding its Codex tool, initially released as a command line tool in April and later transitioning to a web interface.
The latest macOS app for Codex is engineered to enable concurrent operation with multiple agents, combining agent capabilities and cutting-edge workflows. This release closely follows the introduction of GPT-5.2-Codex, OpenAI’s advanced coding model, as the company seeks to attract users of Claude Code.
During a recent press call, CEO Sam Altman highlighted the strength of GPT-5.2 for complex tasks, emphasizing the importance of integrating advanced model capabilities into a more adaptable interface for enhanced usability.
While Altman expressed optimism in GPT-5.2, coding benchmarks present a nuanced perspective. While GPT-5.2 leads on TerminalBench, other agents like Gemini 3 and Claude Opus have posted comparable scores, albeit marginally lower. Results from SWE-bench showcase similar trends, underscoring the competitive landscape of AI’s coding proficiency.
Source: TechCrunch