Google Cloud introduced the Gemini agent at its Gemini at Work event on Oct. 8, describing it as a single workplace agent that can answer questions, create media, write and run code, and complete knowledge-work tasks from one interface. In the company's announcement, Google Cloud CEO Thomas Kurian said people can give the agent objectives instead of a sequence of instructions. The system is designed to plan the work, call skills and tools, connect to business systems, and return completed material in documents, inboxes, and developer environments.
The launch moves Gemini beyond a chat window and into longer-running work. Google says the cloud-based agent can continue a task after someone closes a laptop, while retaining one set of memories and context across devices. It can work inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, as well as through mobile devices, desktop computers, command-line tools, Microsoft 365, and Slack. Reuters independently reported the Workspace, Microsoft 365, and Slack integrations.
A manager for temporary and persistent agents
For jobs with several stages, Google says Gemini can create temporary, job-specific sub-agents and coordinate parallel or sequential steps while tracking the overall objective.
Google is also positioning Gemini as a persistent “coworker agent.” These agents can have a defined organizational role, their own storage, and a dedicated address using the @agents.company.com domain. Google says they receive only the context a user or team provides. TechCrunch reports that a coworker agent gets its own Workspace account and can be tagged, emailed, added to a group chat, or given shared files. Actions appear in an audit trail under the agent's identity rather than a person's.
The agent and the model are separate choices
Gemini is the coordinating agent, but Google says the model handling each job can vary. The system currently routes work among Google's Gemini models and Anthropic's Claude models, selecting a model based on the task. Users can also choose a model, according to TechCrunch. Google says support for additional private and open models is planned.
That separation is part of Google's enterprise pitch. A company could keep its tools, data, skills, and agent context in one system while changing the model used for a particular job. Google also says the product includes spending controls, permissions, identity policies, sandboxing, and network gateways.
The first focus is business customers. TechCrunch reported that Google plans to bring the agent to enterprises before a later consumer rollout. Reuters placed the announcement in a growing competition among major technology companies to sell agents that can act across applications, rather than only respond inside a chat interface.
