Quick Read
- OpenAI launched the standalone Codex app for macOS on February 2, 2026.
- The app enables multi-agent collaboration, allowing AI agents to run tasks in parallel and manage complex workflows.
- New features include ‘Skills’ for integrating external tools and ‘Automations’ for scheduling background tasks.
- OpenAI CEO Sam Altman called it their “most loved internal product.”
- The app runs on the GPT-5.2-Codex model and is available to most ChatGPT subscribers, with plans for a Windows version.
CUPERTINO (Azat TV) – OpenAI officially launched a standalone desktop application for its artificial intelligence-powered coding assistant, Codex, on February 2, 2026. The new macOS app, available to ChatGPT users on Apple devices, signals a significant strategic shift in the competitive AI programming landscape, moving beyond mere code completion to embrace multi-agent collaboration for complex software development tasks.
The release is widely seen as a ‘defensive counter-attack’ by OpenAI in the fiercely contested coding tools sector, where rivals like Google Gemini 3 and Anthropic Claude Opus are rapidly gaining ground. OpenAI CEO Sam Altman lauded the Codex app as “the most loved internal product we’ve ever had,” emphasizing its transformative impact on internal development processes.
New Era of AI-Powered Development with Codex App
The Codex App distinguishes itself from previous web-based or command-line versions by offering a dedicated workspace designed for unified management of multiple AI agents. This new architecture allows developers to assign several coding tasks simultaneously, with agents running independently in the background and executing tasks in parallel. OpenAI describes the app as a ‘command center for agents,’ where results can be reviewed at critical junctures.
Each Codex agent is capable of continuous work for up to 30 minutes, ultimately returning a complete code result. The system currently operates on the GPT-5.2-Codex model, which has achieved top rankings in the TerminalBench benchmark test. Developers can switch between different tasks without losing context, review agent-submitted changes, comment on diffs, and even open code in a local editor for manual adjustments. The app also integrates built-in support for git worktree, enabling multiple agents to work concurrently in the same repository without conflicts, as each operates in an isolated copy of the code.
Beyond Code Generation: Skills and Automation
A core innovation of the Codex App is its expansion of AI coding capabilities beyond simple code generation through the introduction of ‘Skills’ and ‘Automations.’ Skills are reusable packages of instructions, resources, and scripts that allow Codex to call external tools, execute complete workflows, and adhere to team-established development specifications. OpenAI has provided a library of common skills, ranging from obtaining design context from Figma and managing projects in Linear, to deploying web applications to Cloudflare or Vercel, and even generating images and creating documents.
To illustrate the system’s advanced capabilities, OpenAI demonstrated Codex independently developing a racing game, complete with multiple characters, eight maps, and a prop system. This project involved Codex utilizing image generation and web game development skills, consuming over 7 million tokens from initial prompt to full implementation, effectively playing the roles of designer, developer, and QA tester.
The ‘Automations’ mechanism allows developers to schedule tasks for Codex, enabling agents to run continuously in the background. Completed automated tasks are then placed in a review queue for developer approval. Thibault Sottiaux, head of the Codex team, noted that this feature is extensively used within OpenAI for repetitive yet crucial tasks such as daily issue triage, summarizing CI failures, generating release briefings, and routine bug checks.
OpenAI’s Strategy in a Competitive Market
The launch of the Codex App is a strategic move by OpenAI to solidify its position in the rapidly evolving AI programming market. The company aims to make Codex the default tool in this domain, especially given the intensifying competition from other advanced AI coding agents. Data indicates that over 1 million developers used Codex in the past month, with usage nearly doubling since the release of GPT-5.2.
A significant value proposition of Codex, as highlighted by Sam Altman, is its ability to tackle technical debt—tasks human engineers often find tedious, such as refactoring code, cleaning up legacy issues, and improving test coverage. Altman quoted a colleague, saying, ‘Unlike humans, AI coding colleagues won’t run out of dopamine. It won’t stop due to boredom or frustration; it will just keep trying until the problem is solved.’ This continuous, tireless effort is proving invaluable for infrastructure teams grappling with long-standing technical debt within OpenAI.
OpenAI has also emphasized a ‘security-first’ design principle for the Codex system. Similar to the Codex CLI, the app utilizes a native, open-source, and configurable system-level sandbox. By default, agents have limited access, confined to the current working directory or branch and cached web search results. Higher permissions, such as direct network access, require explicit user authorization, with configurable rules for project-specific automatic execution within controlled parameters.
Availability and Future Outlook
Currently, the Codex App is available on macOS for ChatGPT Plus, Pro, Business, Enterprise, and Edu subscribers, with usage included in their subscriptions. Free and Go users can also access a limited-time experience. OpenAI plans to double the rate limits for all paid plans. Looking ahead, the company intends to launch a Windows version, support cloud-triggered Automations, and continuously enhance model capabilities and inference performance.
The release underscores OpenAI’s commitment to bridging the gap between cutting-edge AI models and real-world development workflows. While the Codex CLI’s internals, including its open-source nature, have been detailed by OpenAI in a series of articles, the standalone app focuses on making the full intelligence of the model more accessible and manageable for developers.
The strategic pivot to a standalone, multi-agent application reflects OpenAI’s recognition that the future of AI in software development lies not just in generating code, but in orchestrating complex, collaborative workflows to address long-standing industry challenges like technical debt and developer efficiency, thereby aiming to redefine the very interaction model between humans and AI in engineering.