Gemini Spark comes to Mac with custom MCP support
By Steven Van ·
The always-on agent adds five app integrations, Tasks and Keep links, and real-time topic tracking, though full access needs Google AI Ultra.
Google is pushing its agentic assistant deeper into the desktop. On July 1, 2026, the company announced that Gemini Spark, its 24/7 personal AI agent, is arriving in the Gemini app for macOS, alongside a wave of updates that make the agent noticeably more capable: custom Model Context Protocol (MCP) support, five new third-party app integrations, connections to Google Tasks and Keep, and real-time tracking of topics like sports, stocks, and news.
A native Mac client for an always-on agent
Until now, Spark lived on the web and mobile. Bringing it to a native macOS app matters because a desktop agent can work across the tools you already keep open all day. Google frames Spark as a proactive, always-on assistant that automates time-consuming tasks so you do not have to babysit it. Positioning it on the Mac also puts Spark head-to-head with rival desktop agents like Claude Desktop and Microsoft Copilot. The catch: the full Spark experience is tied to a Google AI Ultra subscription, and the macOS client is rolling out over the coming weeks rather than all at once.
Custom MCP support and five new integrations
The most consequential change for power users is custom Model Context Protocol support. MCP is an open standard for connecting AI assistants to external tools and data, so instead of waiting for Google to build every connector, you can wire your own apps directly into Spark and shape an assistant tailored to your workflow. On top of that, Google is adding native integrations with Canva, Dropbox, Instacart, OpenTable, and Zillow Rentals. In practice that means Spark can design a flyer in Canva, pull a file from Dropbox, reorder your weekly groceries on Instacart, reserve a table on OpenTable, or book an apartment tour through Zillow, all from a single conversation. These connected apps begin rolling out over the next week on web and mobile, with the macOS app following shortly after.
Brain-dumps become action items
Two smaller connections punch above their weight for creators and busy teams. Spark now links to Google Tasks and Google Keep, so you can dump a messy stream of thoughts into Keep and ask Spark to convert it into structured, actionable items in Tasks. For anyone who captures ideas faster than they can organize them, this turns the agent into a lightweight planning layer that sits on top of Google's existing note and to-do apps rather than replacing them.
Real-time tracking of the topics you care about
Finally, Spark can now track topics and react to events in real time. Google highlights use cases like following sports scores, watching stock movements, and monitoring breaking news, but the same capability extends to social media, blogs, online shopping, and weather. Because Spark runs continuously, it can surface changes as they happen instead of only answering when prompted, which is the clearest signal yet of Google's shift from a reactive chatbot toward a genuinely agentic assistant.
Sources: Google Blog, TechCrunch, 9to5Google.