ChatGPT for macOS can now collaborate with popular code editors, terminals, and text editors, allowing users to analyze and edit text directly within these apps.
OpenAI has expanded ChatGPT for macOS with the capability to work directly within applications, starting with code editors, terminals, and word processors. The new functionality is available from version 1.2025.057.
Users can activate the feature via the ChatGPT Chat Bar by pressing Option + Space or through the ChatGPT menu bar icon. ChatGPT can currently work with a range of popular development and text editing apps, including Xcode, VS Code, JetBrains environments, Apple Notes, and Terminal. More compatible apps are expected to follow in the future.
When a compatible app is active, a banner appears in the Chat Bar to indicate which content is being recognized. ChatGPT can analyze the full content of an open editor or terminal, with a limit on the amount of text processed. Users can also make a specific text selection, which ChatGPT will then focus on.
For terminals, such as Terminal, iTerm, and Warp, the last 200 lines of text from the active window are included by default. In code editors and text processors, the entire open file text is included, unless the user makes a selection.
Code Editing and Privacy
For developers, ChatGPT now offers the ability to edit code directly in an IDE without copying and pasting. When a change request is made, ChatGPT generates a diff, allowing users to review and apply the suggested changes. Changes can also be easily undone with CMD + Z.
To work with apps, ChatGPT uses the macOS Accessibility API. Users can disable the feature through macOS settings or the ChatGPT app. For VS Code, a separate extension is required.
Regarding privacy, the shared content remains part of the chat history unless the user deletes it. In that case, the information is erased from OpenAI’s systems within 30 days, unless legal or security reasons prevent this. OpenAI states that content from enterprise services, such as ChatGPT Enterprise, is not used for model training.
Users can completely disable the feature in the settings or temporarily avoid it by enabling Temporary Chat. Admins within an organization can centrally manage the option for all workstations.