Google has officially launched Gemini Spark, its personal AI agent, in India, months after first showing off the tool at Google I/O 2026. The agent is designed to run continuously in the background and carry out tasks on a user's behalf across Google Workspace apps, without needing the user to be actively present or even have their device switched on.
Rollout details and who gets access
Google has confirmed the India launch, with Gemini Spark now built on the company's Gemini 3.6 Flash model, an upgrade from the Gemini 3.5 model it ran on when it was first previewed. The agent will be made available to Google AI Pro and Google AI Ultra subscribers in the country over the coming weeks, positioning it as a premium addition to Google's existing AI subscription tiers rather than a free feature.
How Gemini Spark works in practice
According to Google, users can connect Gemini Spark to Workspace tools such as Gmail, Docs, and Sheets without any separate setup process. Once granted access, the agent is designed to work continuously in the background, a state Google describes as functioning even while a user is asleep, has their phone locked, or has shut their laptop entirely, framing it as a round-the-clock personal assistant. Users retain the option to toggle Gemini Spark off altogether or restrict which specific apps and platforms it is allowed to connect to.
Built-in guardrails for high-stakes actions
Google says the agent has been designed to seek explicit user approval before carrying out tasks it classifies as high-stakes, such as spending money or sending an email without prior review. The company has positioned this permission-based structure as central to Gemini Spark's pitch, framing the tool as one that removes the burden of repetitive digital chores while still keeping consequential decisions in the user's hands.
What Gemini Spark can actually do
Google has outlined several practical use cases for the agent. It can automatically add flight and hotel booking details to a user's travel itinerary as soon as reservations are confirmed, or scan for local events matching a user's interests and log the details into a Google Doc without manual input. The agent is also capable of pulling relevant information from third-party apps, sending users alerts on topics they care about, and autonomously summarising meeting notes or lengthy email threads.
From I/O preview to India launch
Gemini Spark was first unveiled in May during Google's I/O keynote, where the company had said the agent would roll out to users "later this year." Since that preview, the underlying model powering the agent has already been upgraded from Gemini 3.5 to Gemini 3.6 Flash, suggesting Google has continued refining Spark's capabilities in the months leading up to its India debut.
