Google is preparing a new Gemini AI feature that will allow Android users to remotely control their Macs.
Howard Bouchevereau on Unsplash
Key Takeaways
- Google is preparing a new feature that will allow Android users to remotely command and monitor automated AI workflows on their Mac, without needing to be at their desk
- The feature is tied to Gemini Spark, Google’s 24/7 AI agent, and relies on an isolated “new thread” system to prevent data leaking into disconnected sessions.
- The upgrade targets a unique cross-platform ecosystem, allowing Android owners to leverage Google’s AI agent on Apple-silicon Macs ahead of any similar native macOS Copilot feature.
Mac users may soon be able to trigger and monitor agentic AI tasks on their Macs remotely, using the Android Google app.
What Is The Gemini Remote Control Feature For Mac?
Developer Polodarb first spotted a Gemini Spark-related “device picker” string in the Android Google App back in June, but a recent report by Android Authority connects the feature to a new Mac remote-control upgrade in the latest version (v17.36.12). Now, a deeper dive into the code reveals how the feature might prevent data stored on your Mac from leaking..
Relevant text strings include:
- assistant_robin_remote_device_upsell_body “Use your phone to run tasks on your Mac. Download at gemini.google/mac/”
- Assistant_robin_remote_device_disabled “You’ve disabled the Mac”
- Assistant_robin_remote_device_last_seen “Mac last seen”
- Assistant_robin_remote_device_not_available “Your Mac is not available”
- Assistant_robin_device_picker_title “Select a computer”
- Assistant_robin_device_picker_subtitle “Make sure your Mac is online”
- Assistant_robin_device_picker_new_thread_dialog_body_has_device “To work on tasks that aren’t connected to this Mac, start a new thread”
- Assistant_robin_device_picker_new_thread_dialog_body_no_device “To work on tasks that are connected to your Mac, start a new thread”
These code strings are linked in the code to a feature called “Robin,” Google’s internal codename for its Gemini Spark 24/7 autonomous agent.
To use the feature, Gemini for macOS must already be installed on an Apple-silicon-powered device, where it provides on-demand assistance via the Option+Space keyboard shortcut and can view user-selected desktop windows.
We can infer from the text in the code that Gemini’s new remote features will enable a number of interesting usage scenarios, such as allowing power users to trigger and monitor multi-step background automation tasks on their Macs while away from their desks. Gemini already offers integration with Google cloud apps like Docs and Sheets, but this new feature could allow the AI to interact in a similar way with data on your personal Mac.
Alternatively, you could perhaps instruct Gemini to summarize a folder of PDFs on your Mac and send the result to your phone, or set off complex compile scripts with full access to your Mac’s local files and resources.
Such a feature would neatly get around the need for clumsy remote control sessions to interact with AI agents running on a distant Mac.
The “start a new thread” strings above are particularly interesting because they imply that sessions connected to your Mac are kept separate from those that aren’t. This is presumably to prevent data stored on your Mac from leaking into disconnected sessions or onto other Macs in your roster.
Google’s Remote Gemini Feature Is A Mac Exclusive
Perhaps ironically, this function gives Mac owners a head start against Windows users who don’t currently have a standalone Gemini desktop client and must rely on Microsoft Copilot and web apps instead. Even more bizarrely, the feature appears to target that relatively small cohort of Mac owners that opt for Android devices over iPhones.
With no system-level Copilot equivalent built directly into macOS and Apple Intelligence lacking remote background agentic capabilities, Google has a clear path to swoop in and claim this vacuum for Gemini and the Android ecosystem. However, its underlying Gemini Spark AI paradigm, announced this year at I/O 2026, is still highly experimental and its success on macOS remains to be proven.
Leave a comment