Google Cloud launches persistent Gemini Agents with dedicated storage

Google Cloud announced on Thursday that it will launch a new class of AI agents built on its Gemini model, called persistent Gemini Agents. These agents can work on tasks that last hours or days, and each one gets its own Gmail, Calendar, and Drive space to keep its identity and context consistent across applications and devices.

The move comes as businesses race to automate routine work with generative AI. Last year, Microsoft released Copilot for Office, and Google has been developing its own AI tools for Workspace. Persistent memory is the missing piece that lets an agent remember a user’s preferences, past decisions and project details over time, instead of treating each interaction as a blank slate. By tying an agent to its own email, calendar and file storage, Google gives it a “digital persona” that can schedule meetings, draft documents and pull up code snippets without asking the user to re‑enter information. The new agents can also coordinate other agents, turning a single request into a chain of specialized tasks.

Google said the Gemini agents will be available to enterprises through its Cloud AI platform and will integrate with Workspace APIs. Early adopters can use them to build custom workflows, such as automatically drafting a project proposal, scheduling a kickoff meeting and generating a code skeleton for a new feature. The agents will retain a private memory that can be queried through a simple API, and they will be governed by the same security and compliance controls that protect Google Workspace data. Industry analysts expect the feature to reduce manual effort by up to 30% for routine tasks, and to free up engineers to focus on higher‑value work.

The company will roll out the agents in a closed beta next month, with a public preview planned for Q3. Google is also working on a developer portal so that third‑party builders can create agent extensions. The announcement follows a broader push by Google to make generative AI a core part of its cloud offering, after a 2023 investment of over $4 billion in AI research. As the technology matures, Google will need to address privacy concerns around persistent memory, but the company says it will enforce strict data‑access policies and provide audit logs for enterprise customers.