Google Cloud Unveils 'Gemini Agent' for Work, Routing Tasks Across Gemini and Anthropic's Claude

GOOGL's Google Cloud announced Gemini agent on October 8, 2026, describing it as a single, universal agent for work that users give objectives rather than instructions. The agent runs each job on the model that fits best, which today means Google's Gemini family and Anthropic's Claude models, with private and open-weight models planned later. It can split work across temporary groups of specialized sub-agents, running parallel and sequential steps that last hours or days, and keeps session, semantic, procedural and episodic memory in the cloud.

The agent is built to sit inside existing workflows rather than in a separate app. It works through the Gemini Enterprise app and Workspace tools such as Gmail, Docs, Sheets, Slides and Chat, and also inside Slack, Microsoft 365, the command line and an API. Persistent coworker configurations can get their own Workspace account, email address, calendar and Drive storage. Google says it will be included at no additional charge for customers where Gemini Enterprise is available; how persistent coworker accounts and long autonomous tasks will be billed is not yet clear.

Despite some headlines calling it ready for enterprises, Gemini agent is in private preview, with wider availability for select Business and Enterprise Workspace plans promised soon but no date. The launch extends a strong run for Google Cloud, which grew revenue 82% to $24.77 billion in the June 2026 quarter while operating income tripled to $8.81 billion.

What to watch: the general availability date and pricing for long-running tasks, how many enterprises move from preview to paid Gemini Enterprise seats, and how Microsoft responds to an agent designed to work inside Microsoft 365.