- Type
- Always-on cloud AI agent
- Developer
- Google (Google DeepMind / Gemini app)
- Announced
- 19 May 2026, Google I/O
- Built on
- Gemini 3.5 on the Google Antigravity platform
- Execution
- Dedicated Google Cloud virtual machine
- Availability
- Google AI Pro ($19.99/mo) and AI Ultra, 160+ countries since 30 July 2026
- Related
- OpenAI Codex, Claude Cowork, Always-on agents
- Type
- Always-on cloud AI agent
- Developer
- Google (Google DeepMind / Gemini app)
- Announced
- 19 May 2026, Google I/O
- Built on
- Gemini 3.5 on the Google Antigravity platform
- Execution
- Dedicated Google Cloud virtual machine
- Availability
- Google AI Pro ($19.99/mo) and AI Ultra, 160+ countries since 30 July 2026
- Related
- OpenAI Codex, Claude Cowork, Always-on agents
History
Spark was announced on 19 May 2026 during the Google I/O keynote, described by Google CEO Sundar Pichai as "your personal AI agent in Gemini app that helps you navigate your digital life, taking action on your behalf and under your direction". It was the first always-on personal agent shipped by a major AI laboratory: at launch, trusted testers received access the same morning, with the wider beta opening the week of 25 May 2026 for Google AI Ultra subscribers in the United States.[2][3]
Availability broadened quickly. From 30 July 2026 Spark was included in the $19.99 per month Google AI Pro tier — previously limited to the $99.99–$199.99 AI Ultra plans — across the United States and more than 160 other countries; Google AI Plus, the free tier, and work or school accounts do not include it.[1][4]
Technology
Spark is built on Gemini 3.5 and Google's Antigravity agentic platform — the same harness underlying Google's developer agents. Rather than running on the user's device, it executes on a persistent cloud virtual machine, which is what distinguishes it from session-bound assistants: tasks continue after a laptop closes or a phone locks, and the agent can schedule work, monitor changes and act when something happens.[2][3]
Its action surface covers first-party Google Workspace applications — Gmail, Calendar, Docs, Drive — plus third-party services connected through the Model Context Protocol (MCP) at launch (Canva, OpenTable and Instacart among them). Users can message Spark directly by text or email, create custom sub-agents for recurring jobs, and grant spending authority with explicit budgets and merchant limits — a design choice intended to make autonomous action auditable rather than open-ended.[2][4]
The persistent-VM model also shifts risk: an agent with continuous access to email and payment authorisation expands the attack and error surface, which Google addressed through direction-under-user-control framing, budget caps and staged rollout.
Applications and Impact
Reported uses include inbox triage and drafting, meeting and calendar management, ongoing research threads that resume without prompting, shopping and booking tasks executed inside connected apps, and multi-step errands that span several services. More broadly, Spark popularised the always-on agent category — a category that did not exist in mainstream products six months before its launch — and reframed the assistant market around delegation and continuity rather than per-session chat, pushing competitors toward persistent, cloud-hosted agents of their own.[1][2]
>See Also
🇲🇾 Relevance to Malaysia: Spark reached Malaysia in the 30 July 2026 expansion to 160+ countries, making it one of the first always-on agents from a major lab available locally, at the $19.99 (about RM 84) Google AI Pro price point — a subscription Malaysian consumers and small businesses can evaluate without enterprise contracts. Its value proposition suits Malaysia's mobile-first, SME-heavy market: solo founders and small teams can delegate inbox, scheduling and document work without hiring support staff, complementing the AI-assisted productivity that MDEC promotes through its digitalisation programmes. Two cautions apply. First, data handling: agent access to Gmail, Drive and Calendar means personal and company data processed on Google's servers, engaging the cross-border transfer provisions of the Personal Data Protection Act and the data-governance expectations in the National AI Office's AI framework — work accounts under Google Workspace policies should follow organisational rules rather than personal subscriptions. Second, autonomous spending features make budget caps, merchant limits and per-transaction review essential before granting payment authority in a Malaysian business context.
References
- ↑Fello AI. Best AI Models in November 2026 (Gemini Spark). https://felloai.com/best-ai-models/
- ↑Google. The Gemini app becomes more agentic, delivering proactive, 24/7 help (I/O 2026). https://blog.google/innovation-and-ai/products/gemini-app/next-evolution-gemini-app/
- ↑Digital Applied. Gemini Spark: Google's Always-On Personal AI Agent. https://www.digitalapplied.com/blog/gemini-spark-always-on-personal-agent
- ↑Harvard Business School AI Institute. Gemini Spark and the Rise of the Always-On Agent (July 2026). https://aiinstitute.hbs.edu/future-proof-with-ai/memo-july-2026/