Google Cloud introduces the Gemini agent.

Today at Gemini at Work 2026, Google Cloud announced the Gemini agent, a universal agent for work.

With Gemini, work now starts in the prompt window:

Gemini has all of an organization’s business context and can be used for everything from knowledge work to answering questions, and content creation to coding, all from a single prompt box.
It plans the work, uses skills and tools, connects to Cloud customers’ business systems, and brings back something finished — inside the documents, the inbox, and the developer environments they already work in.
It chooses the best model for the job, has built-in cost controls, and most importantly, has the security, administration, and governance required by enterprise customers.

Check out the Google Cloud blog to learn more.

Text "Google Cloud Gemini at Work '26" on colorful bubbles with a spark and arrow

Google Cloud has introduced the Gemini agent, a single, unified “universal agent for work” designed to automate complex, multi-step enterprise workflows. Unveiled at the Gemini at Work 2026 conference, the agent shifts the paradigm from simple prompting to persistent execution, allowing users to give it overarching objectives rather than granular instructions. Powered by Google’s latest Gemini 4 Argon frontier model, it can maintain deep context over hours or days to return completed background deliverables. [1, 2, 3, 4, 5, 6]

Core Capabilities & Architecture

Multi-Agent Orchestration: The platform coordinates dedicated sub-agents (e.g., CRM, Web Search, Compliance, and Creative Review agents) to execute complex, multi-phase projects independently. [1]
Persistent Execution: It runs 24/7 in the cloud. Users can assign a task, close their devices, and return later to a finished product in their email or document space. [1, 2, 3]
Multi-Model Routing: To balance performance and efficiency, the system automatically routes tasks dynamically—leveraging both Google’s Gemini models and Anthropic’s Claude models. [1, 2]
Omnipresent Access: It operates across the web, iOS, Android, macOS, and Windows, integrating seamlessly into Google Workspace, Microsoft 365, Slack, and developer command lines. [1, 2]

Creating “Coworker Agents”

Beyond a personal digital assistant, businesses can provision persistent Coworker Agents. These digital entities are treated like real team members, coming equipped with their own: [1, 2]

Workspace account and dedicated email address
Secure Google Drive storage and corporate directory presence
Ability to be @mentioned directly inside a Chat space or Google Doc comment to delegate work [1]

Enterprise Governance & Security Infrastructure

To prevent “agent sprawl” and mitigate risks like data exfiltration, Google simultaneously launched a specialized Agent Enterprise Platform featuring advanced governance tools: [1, 2]

Agent Compute Engine: A purpose-built, hardware-isolated runtime sandbox environment designed to securely execute agent sessions.
Agent Gateway & Registry: Functions as a centralized firewall and air traffic controller to log, audit, and apply strict semantic policies to all agent traffic and tool access.
Agent Identity: Assigns cryptographically attested, unique IDs to every agent using the SPIFFE standard for rigorous auditing and access control. [1, 2]

Availability

The Gemini agent is available across the Google Cloud Platform. Dedicated previews are currently live for Financial Services and Legal specializations, with tailored rollouts for government (supporting FedRAMP High boundaries), healthcare, and retail sectors underway. [1, 2, 3]

If you would like to explore this further, let me know if you want to:

See an example of how to set up a Coworker Agent for a specific team role.
Learn more about the underlying capabilities of the Gemini 4 Argon model.
Dive deeper into the security controls like Agent Gateway or SPIFFE identities.