Google Cloud Launches the Gemini Agent, a Work Agent That Gets Its Own Email Address

Google Cloud on October 8, 2026 announced the Gemini agent, a single agent for work that answers questions, handles knowledge work, creates images and media, and writes and runs code from one prompt box and one API. Coworker agents built with it get their own Workspace account, email address and calendar, and act under their own identity rather than a person's, the company said.

Google Cloud CEO Thomas Kurian presented the agent in his Gemini at Work 2026 keynote. It follows Microsoft's updated Copilot on September 25 and OpenAI's Dots agents on September 29 (Implicator), as the big vendors move AI agents, software that plans and carries out multi step tasks on its own, into everyday office work. Google CEO Sundar Pichai said at the event that Gemini has more than 1 billion monthly active users, and Google will bring the agent to businesses before consumers (TechCrunch).

What did Google announce at Gemini at Work 2026?

Google announced the Gemini agent on October 8, 2026 as one agent that combines chat, autonomous task work and coding, which users give "objectives, not instructions," in Kurian's words. Users can assign it work, schedule tasks or have it respond to events, and it can act as a personal assistant or on behalf of a team or a role, such as an analyst in a finance department.

The agent runs in the cloud, so work that takes hours or days continues after the user closes a laptop, and it keeps one set of memories across devices, according to Google. It is reachable on the web, iOS, Android, Windows and Mac, from the command line, and inside Google Workspace, Microsoft 365 and Slack. For larger jobs it creates temporary sub-agents and coordinates them.

Google says the agent keeps four kinds of memory: for the current task, for knowledge it gathers, for how jobs get done and for what it did before. It connects to tools such as Salesforce, ServiceNow, Jira, BigQuery, Snowflake and Databricks, and to any Model Context Protocol (MCP) server, the open standard for linking agents to outside tools and data.

How do coworker agents work in Google Workspace?

A coworker agent is a Gemini agent with a fixed role on a team: Google says users describe the role, and Gemini creates an agent with its own Workspace account, email address, calendar, Drive and entry in the company directory. Colleagues add it to a Chat space or @mention it, and its edits in Docs appear under its own name in version history.

Google says the addresses use an @agents.company.com format, that a coworker agent sees only what people share with it, and that access follows existing sharing and membership. Every action goes into an audit trail attributed to the agent rather than to a person. Users follow the agent's work, including its delegation to sub-agents, in a "tasks inbox" (TechCrunch).

Which models does the Gemini agent use?

Google says the Gemini agent picks a model for each job, routing work across its Gemini models and Anthropic's Claude models today, with other private and open models to follow. Users can also choose a model themselves (TechCrunch), and administrators can restrict which models agents may use (VentureBeat).

Google says prompts, documents and chat histories stay inside Gemini Enterprise when Claude is used, but it has not said whether that data goes to Anthropic for processing or where it is kept (VentureBeat). The Claude option arrives two days after Anthropic opened its own route into the same apps: on October 6, Anthropic put a Claude add-on for Google Docs, Sheets and Slides into public beta on its paid plans. Workspace customers can now reach Claude through Google's agent or through Anthropic's sidebar.

How does Google control what the agent does and spends?

Google says every agent gets a cryptographically attested identity with least privilege permissions approved by security administrators, runs code inside an isolated Agent Sandbox, and sends all traffic through Agent Gateway, which Google calls an AI network firewall. Administrators write a policy once, such as a ban on opening documents classified Need to Know, and it applies to every agent.

On cost, a Smart Routing tool sends each task to the model that Google says gives the best performance at the lowest cost, and administrators can set a hard spend cap per project in the Cloud Billing Console. When a cap triggers, that project's agent pauses until someone resumes it. The caps themselves are not new: Google introduced project level spend limits for agent workloads on August 26, 2026. Cohere's North 2, launched on October 5, sets quotas and caps down to the individual user and agent, while Google tracks spend per project so costs can be charged back to departments.

Who can use the Gemini agent, and when?

The Gemini agent is in private preview for selected customers, and Google says it will be included at no additional charge wherever Gemini Enterprise is available, including existing Workspace business and enterprise subscriptions (VentureBeat). Google has given no general availability date and says more pricing details will follow closer to the wider release.

Google executives told industry analysts after the event to expect general availability around the end of October or early November (Implicator). Versions for financial services and legal work are in preview, and versions for government, healthcare and retail are coming, according to Google. Google also says nearly 90% of the Fortune 100 use Gemini Enterprise, a figure that covers the existing product, not the new agent.

What we don't know yet

• When the Gemini agent becomes generally available, and in which regions, including the EU.
• Whether coworker agents need their own Workspace licenses, and whether long running tasks and sub-agents carry usage charges.
• Whether data sent to Claude models is processed by Anthropic, and under which retention rules.
• How administrators revoke a coworker agent's access and identity.
• How reliably the agent completes multi day tasks; Google has not published independent tests.