Gemini Spark Goes Live: Google’s 24/7 AI Agent Lands for US Users
3 min readGoogle flipped the switch on broader US access to Gemini Spark this week, opening its cloud-resident “24/7” personal AI agent to all Google AI Ultra subscribers in the United States and adding a wave of third-party app integrations that pull the agent out of the Workspace bubble for the first time.
Spark was unveiled at Google I/O on May 19 as part of a sweeping update to the Gemini app, but it shipped that day only to a small group of trusted testers. Today marks the start of its general US rollout to paying Ultra subscribers, alongside new Model Context Protocol connections that let Spark take action inside Canva, OpenTable, and Instacart.
What Gemini Spark Actually Does
Unlike a chat assistant that goes idle when you close the tab, Spark runs on dedicated Google Cloud virtual machines around the clock. That means it can keep working on multi-step tasks while your laptop is shut: drafting follow-up emails from a Gmail thread, monitoring price drops on a flight, organizing receipts in Drive, or queueing a grocery order for the weekend.
The agent natively reasons across Gmail, Calendar, Drive, Docs, Sheets, Slides, YouTube, and Maps. With the MCP additions announced today, it now extends into design (Canva), restaurant reservations (OpenTable), and groceries (Instacart), with more partners staged for summer. According to TechCrunch, the rollout is being staged through the Google AI Ultra plan, which Google cut from $249.99 to $99.99 per month at I/O to make agent access cheaper.
The Race Behind the Launch
Spark is Google’s most direct answer to OpenAI’s Operator and Anthropic’s computer-use agents. Where Operator drives a remote browser and Anthropic’s tooling controls a desktop, Spark sits closer to a personal assistant model: an always-on background worker bound to a user’s identity, with explicit approval gates before sensitive actions like sending messages or placing orders.
The MCP-based integration approach is also notable. By plugging in third-party apps through the open Model Context Protocol rather than bespoke APIs, Google is signaling that the same plumbing it already supports across the Gemini ecosystem will be the on-ramp for any developer who wants their app to be reachable by an agent.
Why It Matters
This is the first time a major lab is offering a persistent, cloud-resident AI agent to a mass consumer audience at a sub-$100 price point. If Spark behaves well in the wild, it sets a new baseline expectation: AI assistants do not just answer questions in real time, they take initiative on your behalf while you sleep.
The big watch items are reliability and trust. Autonomous agents that touch email, calendars, and payment-adjacent apps have a much smaller error budget than chatbots. International users will also be waiting: the EU rollout is held up by AI Act conformity assessments, with Q3 2026 the earliest realistic timeline.