IT之家 · 智能时代 · 10/8/2026, 21:21:00
Google Cloud Launches Gemini Agent: A Multi-Model General Work Agent Enters Private Preview
Google Cloud unveiled Gemini Agent, a 'general work agent' currently in private preview, during the Gemini at Work 2026 event. The tool enables users to delegate tasks like knowledge work, media creation, and coding by simply setting goals, integrating with Workspace and third-party services. Notably, it supports both Gemini and Claude models under the hood, with plans for broader model compatibility and developer APIs.
SOURCE COVERAGEOriginal coverage
On October 8, Google Cloud announced the launch of Gemini Agent for enterprise customers at its "Gemini at Work 2026" event. Positioned as a "general-purpose work agent," this product supports Gemini Enterprise, Workspace, and third-party services.
According to Google, Gemini Agent only requires users to specify their goals to complete tasks. It supports answering questions, handling knowledge-based work, creating media content, and writing code.
For example, if your manager emails you asking for a presentation on recent project progress, Workspace Intelligence identifies that the task can be handled by AI. Users simply need to click a button to hand it over to Gemini.
Beyond acting as a personal assistant, Gemini Agent can also serve as a "team member." It can take on roles such as Project Manager or Financial Analyst within a team.
IT Home noted that the agent currently supports Gemini and Claude models, with plans to become compatible with various closed-source and open-source models in the future. Google will provide an API allowing developers to integrate Gemini Agent into third-party applications.
Gemini Agent is currently in private preview and is expected to roll out broadly to eligible Workspace customers soon, supporting select Business and Enterprise plans.