Google Spark AI agent for macOS

The macOS launch

The headline change this week is that Spark can now reach beyond the chat window and act directly on your Mac’s files and apps. Instead of manually sorting a Downloads folder, you ask Spark to sort every PDF into the right place, and it does it. It also bridges your desktop and Google Workspace, so a request like “build a budget spreadsheet from the invoices on my computer, and keep it updated” becomes a single instruction rather than an afternoon of admin. Access to your files is permission-based, meaning that Spark can only touch what you’ve explicitly allowed it to see.

The more striking promise is remote execution, which Google says is coming soon rather than going live today. Using it, you’ll be able to hand Spark a multi-step task from your phone, find a specific sales report on the Mac, pull the total revenue figure, email it to you, and have it complete the job on your computer while you’re out. It’s the kind of feature that makes the “agent” label feel earned rather than marketing.

For now, Gemini Spark for macOS is in beta, limited to Google AI Ultra subscribers aged 18 and over, and limited to the US. It’s a download away at gemini.google/mac.

More apps, and reacting for you

Spark’s connected-app list has grown considerably. It already works with Gmail, Calendar, Drive, Docs, Sheets, Slides, YouTube and Maps. This update adds Google Tasks and Google Keep, so a scattered pile of Keep notes can be turned into an organised Tasks list on request. Beyond Google’s own ecosystem, Spark now integrates with Canva, Dropbox, Instacart, OpenTable and Zillow Rentals, covering everything from knocking out a quick flyer to booking a dinner reservation or browsing a rental listing.

Google is also opening up custom Model Context Protocol (MCP) support, letting you wire your own preferred apps into Spark rather than waiting for Google to build an official integration. These connected-app updates are rolling out over the next week on web and mobile, with the macOS app catching up in the following weeks.

The other addition is real-time topic tracking. Rather than refreshing a live blog after your team’s match finishes, you can ask Spark to notify you the moment full-time analysis lands. Set a price threshold on a stock, and Spark will send a financial report the instant it’s hit. Google says Spark can now monitor: 

Effectively, it turns what you’re monitoring into a standing watch list that pings you rather than the other way round.

Where this sits in the bigger picture

Spark is Google’s answer to a race that every major AI lab is currently running, and is most useful to people already living within Google’s ecosystem. Google has said more Spark features are coming this summer, possibly even the ability to spend money on your behalf, giving you another bullet in your arsenal of AI tools.