Google announced plans to integrate a new “Spark” agent and enhanced voice control into its Gemini app for macOS this summer. These features were previewed at I/O 2026.
The additions aim to significantly expand the Gemini app’s capabilities on Apple’s desktop platform, offering users more direct interaction methods and advanced workflow automation.
The “Spark” agent is described as a 24/7 personal AI agent designed to manage tasks and navigate digital life on a user’s behalf. It integrates with Google Workspace applications such as Gmail and Docs, along with third-party services. For macOS, Spark will specifically enable tasks involving local files and automate workflows across the desktop. This includes the ability to use any open windows as context for prompts.
Initially, Gemini Spark will be available in beta to Google AI Ultra subscribers in the U.S. This rollout is expected to begin next week for Android, iOS, and web versions of the Gemini app, with macOS integration following this summer.
The new voice experience will allow for more natural, free-flowing speech, eliminating the need to carefully articulate every word. Gemini will process these spoken inputs, automatically reformatting them into precise drafts and capturing the user’s intent. The feature leverages context from the user’s screen.
During the I/O 2026 demonstration, Google showed the voice control being used to select files in Finder and then dictate an email that was automatically inserted into a Gmail compose window. Users will activate this feature by long-pressing the function key on their Mac, which displays a floating prompt pill. This voice experience will be released globally to all users.
Google introduced the native Gemini app for macOS in April, with a “small team” utilizing Antigravity for development. The expansion of features on macOS aligns with Google’s strategy to make its AI technology broadly accessible across various operating systems. For more details on Google’s AI initiatives, visit the Google AI Blog. The Gemini app is available for download for all users.
The new “Spark” agent and voice control capabilities are scheduled for release later this summer. Specific rollout dates and a comprehensive list of all functionalities will be provided closer to the official launch.
