OpenAI's Codex just got its own Mac app - and anyone can try it for free now
· ZDNET
OpenAI has launched a dedicated Mac application for its Codex AI coding agent, marking a significant evolution from single-agent coding operations to multi-agent software development workflows. This new application, distinct from the general-purpose ChatGPT app, serves as a command center for managing multiple coding agents across complex, long-running projects and tasks. The timing of this release is particularly noteworthy, coming just weeks after the launch of GPT-5.2-Codex in mid-December. According to OpenAI CEO Sam Altman, this latest model represents a breakthrough in AI capabilities, crossing "a threshold of real utility" that has driven explosive adoption. The company reports that Codex usage has nearly doubled since the December launch, with over one million developers using the platform in the past month alone. Altman described GPT-5.2-Codex as "the fastest adopted model that we have ever made," with usage growing more than 20 times since August. The new Mac app addresses a fundamental limitation that OpenAI identified as AI models became more capable. As Altman explained during a company briefing, "There's obviously been a huge shift in software agents over the last few months as the models have crossed a threshold of real utility. 5.2, in particular, is a model that many of us have found can do extremely complex things, and we realized we started to feel limited by the interface." The application introduces several innovative features designed to support complex development workflows. Central to its functionality is the ability to run multiple tasks in parallel while helping users supervise agents across longer, end-to-end work processes. The app also introduces "skills," similar to those found in Claude Code, which allow developers to define repeatable workflows such as fetching logs, fixing tests, and handling coding-adjacent tasks like summarizing threads and completing tickets. Security has been a primary consideration in the app's design. Running AI agents with write access on local machines presents inherent risks, which OpenAI has addressed through a sophisticated sandbox model. The system limits folder write access and network permissions to only approved directories, with developers able to set granular approval levels including Untrusted, On failure, On request, and Never. This approach allows for powerful local AI assistance while maintaining security boundaries. One of the most compelling aspects of the new application is its seamless integration with existing development workflows. Users can switch between their preferred IDE, terminal, and the new app while maintaining context across all tools. This continuity represents a significant advancement in AI-assisted development, allowing developers to work in whatever environment suits their current task without losing the AI's understanding of their project. The real-world impact of these capabilities is already being demonstrated by OpenAI's own development team. Several developers reported that they've begun working almost entirely within the Mac app, rarely needing to open their traditional IDEs or terminal windows. Altman shared his own experience, describing how he completed "a fairly big project in a few days" without opening an IDE even once, something he "did not think was going to be happening by now." Major enterprise customers are already leveraging these capabilities, with OpenAI reporting adoption by companies including Cisco, Ramp, Virgin Atlantic, Vanta, Duolingo, and Gap. This enterprise traction suggests that AI-assisted coding is moving beyond experimental use cases into production environments at scale. The pricing structure reflects the varying needs of different user segments. While Codex is available with limited usage in the $20-per-month ChatGPT Plus plan, more extensive resource availability comes with the $200-per-month ChatGPT Pro plan. For companies requiring even higher performance, OpenAI is exploring additional tiers focused on computational speed rather than just token allocation, recognizing that enterprise users often prioritize rapid iteration over cost optimization. To encourage broader adoption, OpenAI has announced a limited-time promotion that includes Codex access with ChatGPT Free and Go plans, while doubling rate limits across all paid tiers. This democratization of access could significantly accelerate the adoption of AI-assisted coding across the developer community. The application also features a new "plan mode" that allows for read-only review of code, ensuring the AI won't make unwanted modifications, along with selectable personalities for different interaction styles. Additionally, OpenAI has expanded IDE support with extensions for JetBrains IDEs, addressing developers who prefer these environments over VS Code. Looking ahead, this launch represents more than just a new application—it signals a fundamental shift in how software development might be conducted. OpenAI envisions programmers coordinating teams of agents across all stages of the software lifecycle, from design and build to shipping and ongoing maintenance. This multi-agent approach could dramatically reshape development workflows, potentially allowing smaller teams to accomplish what previously required much larger engineering organizations. The Mac-only initial release follows OpenAI's typical rollout pattern, with a Windows version likely to follow based on the company's previous app launches. As AI coding capabilities continue to advance, tools like the new Codex app may become as essential to developers as traditional IDEs are today, fundamentally changing how software is conceived, built, and maintained.
OpenAI has unveiled a groundbreaking Mac application dedicated to its Codex AI coding agent, representing a pivotal shift from traditional single-agent coding assistance to comprehensive multi-agent software development orchestration. This specialized application, distinct from OpenAI's general-purpose ChatGPT app, functions as a sophisticated command center designed to manage multiple coding agents across complex, extended projects and development workflows.
The launch comes at a remarkable moment in AI development history. Just weeks after introducing GPT-5.2-Codex in mid-December, OpenAI is witnessing unprecedented adoption rates. CEO Sam Altman revealed during a company briefing that Codex usage has nearly doubled since the December launch, with over one million developers actively using the platform in the past month. More striking still, Altman described GPT-5.2-Codex as "the fastest adopted model that we have ever made," with usage metrics showing growth exceeding 20 times since August.
This explosive adoption reflects what Altman characterizes as a fundamental breakthrough in AI capabilities. "There's obviously been a huge shift in software agents over the last few months as the models have crossed a threshold of real utility," he explained. "5.2, in particular, is a model that many of us have found can do extremely complex things, and we realized we started to feel limited by the interface." The new Mac app directly addresses this interface limitation, providing developers with tools capable of matching the sophistication of the underlying AI models.
The application introduces revolutionary features that redefine AI-assisted development. At its core is the ability to execute multiple tasks in parallel while providing comprehensive oversight of agents working on longer, end-to-end development processes. The system incorporates "skills"—similar to those found in Anthropic's Claude Code—that enable developers to create repeatable workflows for common tasks such as log retrieval, test fixing, and coding-adjacent activities like thread summarization and ticket completion.
Security considerations have been paramount in the application's design. The inherent risks of running AI agents with write access on local machines required innovative solutions. OpenAI has implemented a sophisticated sandbox model that restricts folder write access and network permissions to pre-approved directories. Developers can configure granular approval levels including Untrusted, On failure, On request, and Never, ensuring powerful AI assistance while maintaining strict security boundaries.
Perhaps most impressively, the application achieves seamless integration with existing development environments. Users can fluidly transition between their preferred IDE, terminal, and the new app while preserving complete context across all tools. This continuity represents a significant advancement in AI-assisted development, eliminating the friction typically associated with switching between different development environments.
The practical impact of these capabilities is already evident within OpenAI itself. Multiple developers on the team report working almost exclusively within the Mac app, rarely requiring traditional IDEs or terminal windows. Altman shared his personal experience: "I was astonished by this, but I did a fairly big project in a few days earlier this week and over the weekend. I did not open an IDE during the process. Not a single time." He emphasized that this level of AI-powered development capability arrived sooner than anticipated: "I did not think that was going to be happening by now."
Enterprise adoption is already substantial, with major corporations including Cisco, Ramp, Virgin Atlantic, Vanta, Duolingo, and Gap integrating Codex into their development workflows. This enterprise traction indicates that AI-assisted coding has evolved beyond experimental applications into production-ready solutions capable of supporting mission-critical development processes.
The pricing structure reflects diverse user requirements across the developer ecosystem. While Codex access is available with usage limitations in the $20-per-month ChatGPT Plus plan, more extensive resources are provided through the $200-per-month ChatGPT Pro plan. Recognizing enterprise demands for computational speed over cost optimization, OpenAI is developing additional tiers focused on rapid iteration capabilities rather than merely token allocation limits.
To accelerate adoption, OpenAI has announced a limited-time promotion extending Codex access to ChatGPT Free and Go plans while doubling rate limits across all paid tiers. This democratization strategy could significantly broaden AI-assisted coding adoption throughout the global developer community.
Additional features include a "plan mode" enabling read-only code review to prevent unwanted modifications, selectable interaction personalities, and expanded IDE support through JetBrains extensions. This JetBrains integration is particularly significant for developers who prefer these environments over VS Code, addressing a previously underserved segment of the development community.
The application's launch follows OpenAI's established rollout pattern, beginning with Mac-only availability before expanding to Windows. This approach allows for focused optimization and feedback collection before broader platform deployment.
Looking toward the future, this release signals a fundamental transformation in software development methodology. OpenAI envisions developers orchestrating teams of AI agents across the complete software lifecycle—from initial design and development through deployment and ongoing maintenance. This multi-agent paradigm could dramatically reshape development workflows, potentially enabling smaller teams to accomplish what traditionally required much larger engineering organizations.
The implications extend beyond individual productivity gains to fundamental changes in how software companies structure their development processes. As AI agents become capable of handling increasingly complex tasks autonomously, the role of human developers may evolve from direct coding to high-level orchestration and strategic decision-making.
This evolution represents more than technological advancement—it suggests a new era in software development where human creativity and AI capability combine to achieve unprecedented productivity and innovation. The new Codex app may well be remembered as a pivotal moment when AI-assisted development transitioned from helpful tool to essential infrastructure.
Referenced links
- Stop using ChatGPT for everything: My go-to AI models for research, coding, and more
- I got 4 years of product development done in 4 days for $200
- I used GPT-5.2-Codex to find a mystery bug and hosting nightmare
- 10 ChatGPT Codex secrets I only learned after 60 hours of pair programming
- Is ChatGPT Plus still worth your $20?