ChatGPT is moving from code advice into direct code changes. The newest ChatGPT app for macOS can now edit code inside supported developer tools, giving OpenAI a more hands-on role in software workflows.
The feature is available first to ChatGPT Plus, Pro, and Team subscribers who update the macOS app. OpenAI says Enterprise, Edu, and free users will get access next week.
What the new macOS feature does
The updated ChatGPT app for macOS can take action inside supported coding environments. The tools named in the source are Xcode, VS Code, and JetBrains.
That matters because the interaction is no longer limited to asking ChatGPT for code, then copying the answer into an editor. With direct code editing, the app can make changes where the developer is already working.
OpenAI is also offering an optional “auto-apply” mode. When enabled, ChatGPT can make edits without requiring additional clicks from the user.
That setting changes the level of friction in the workflow. A developer can keep more of the exchange inside the development environment, while still deciding whether the extra automation is appropriate for the task at hand.
Who gets access first
The rollout begins with paid and workplace plans. Subscribers to ChatGPT Plus, Pro, and Team can use direct code editing as of Thursday by updating the macOS app.
OpenAI says the feature will roll out to Enterprise, Edu, and free users next week. That creates a staged release, with the first group getting access before broader availability.
The Windows version is not included in the initial launch. In a post on X, Alexander Embiricos, a member of OpenAI’s product staff working on desktop software, said the ChatGPT app for Windows will get direct code editing “soon.”
OpenAI Developers also posted that ChatGPT for macOS can now edit code directly in IDEs and that the feature is available to Plus, Pro, and Team users.
How this builds on “work with apps”
Direct editing follows OpenAI’s earlier “work with apps” capability for ChatGPT on macOS. That feature launched in beta in November 2024.
The earlier version let the ChatGPT app read code inside a handful of developer-focused coding environments. Its main practical effect was to reduce the need to copy and paste code into ChatGPT.
The new feature extends that path. Instead of only reading code and responding with suggestions, ChatGPT can now act on the code in supported tools.
For developers, the distinction is important. Reading context helps an assistant understand what is open in the editor. Editing context turns that assistant into a participant in the code change itself.
Why this puts ChatGPT closer to AI coding tools
With direct code editing, ChatGPT now competes more directly with AI coding tools such as Cursor and GitHub Copilot. Those products are already associated with software development workflows, while ChatGPT has often been used as a more general AI assistant.
The source also notes that OpenAI reportedly has ambitions to launch a dedicated product to support software engineering in the months ahead. Direct editing in the macOS app fits into that broader direction without requiring a separate product today.
AI coding assistants are already widely used. The source cites GitHub’s latest poll, in which the vast majority of respondents said they had adopted AI tools in some form.
It also cites a recent claim from Y Combinator partner Jared Friedman that a quarter of YC’s W25 startup batch have 95% of their codebases generated by AI. That claim points to how far some teams are willing to push AI-generated software, even as the risks remain unsettled.
The risks remain part of the story
The same shift that makes AI coding tools useful also raises concerns. The source names security, copyright, and reliability risks associated with AI-powered assistive coding tools.
A survey from software vendor Harness found that the majority of developers spend more time debugging AI-generated code and security vulnerabilities compared to human-written contributions.
A Google report found a mixed picture. According to the source, AI can speed up code reviews and help documentation, but at the cost of delivery stability.
That makes direct editing a productivity feature with a clear tradeoff. Reducing copy and paste, adding IDE access, and offering “auto-apply” can speed up the mechanics of development. But the output still has to be reviewed, tested, and understood.
For teams adopting ChatGPT code editing on macOS, the practical question is not only whether the assistant can make a change. It is whether the change is secure, reliable, and maintainable once it enters the codebase.