Google has introduced Gemini Spark, a new AI agent that runs on the Gemini 3.5 model and Google's Antigravity system. The company announced the product during its I/O event in 2026. This software tool focuses on deep integration with existing productivity platforms rather than standalone operation.
The new software tool focuses on deep integration with existing productivity platforms rather than standalone operation.
The agent connects directly to Google Workspace applications through Model Context Protocol (MCP). It also establishes new integrations with Canva, OpenTable, and Instacart via the same protocol. The system maintains functionality by continuing to process tasks in the background even when users disconnect from the internet. Future updates will add direct texting and emailing capabilities along with custom sub-agents.
Google is rolling out Gemini Spark to trusted testers this week in the United States. The beta program for Google AI Ultra subscribers begins next week in the same region. The company has not confirmed exact dates beyond these initial rollout windows.
The product represents a shift toward autonomous task execution within established software ecosystems. It builds upon previous iterations by adding offline processing and broader third-party service connections through MCP. This approach allows the agent to operate across multiple platforms without requiring users to switch between applications.
Source: GIZMOCHINA
