Google launches Gemini Spark AI for macOS desktop agent

Google has released Gemini Spark for macOS in a limited beta, turning the Gemini AI from a conversational chatbot into a desktop‑agent that can carry out multi‑step tasks on a user's computer. The update lets the AI organise files, extract information from PDFs, create spreadsheets from invoices and summarise documents, all while operating on folders the user explicitly shares. Spark also monitors topics over time and can notify users of changes, expanding its role beyond single‑prompt responses.

The macOS version integrates with a growing list of third‑party services, including Google Keep, Google Tasks, Canva, Dropbox, Instacart, OpenTable and Zillow Rentals, and uses the Model Context Protocol (MCP) to allow developers to add further tools. A future feature will let users start jobs from a phone that continue to run on the Mac after they walk away. The beta is currently available only to Google AI Ultra subscribers in the United States, positioning the service against competitors such as Microsoft Copilot and Claude Desktop.

Overall, Gemini Spark’s rollout marks Google’s shift toward “agentic” AI that automates repetitive desktop workflows rather than merely answering queries.