Google Cloud announced the Gemini agent on October 8 at its Gemini at Work event. It is one agent for questions, documents, images and code, and it works in Google Workspace, Microsoft 365 and Slack. It picks a model for each task, from Google’s Gemini or Anthropic’s Claude. Versions for finance and legal work are in preview.
New to this? Read it in simple words
- Google made one AI agent for many kinds of office work.
- It can keep working on a task for hours or days.
- It can use Google’s Gemini or Anthropic’s Claude models.
- Google has not given a price or a date for most customers.
- AI agent
- An AI that takes actions to reach a goal, not only answers questions.
- MCP server
- A program that lets AI agents use a tool or a data source in a standard way.
- Preview
- An early version that only some customers can use before the full launch.
What it does
Google calls it “a single, universal agent for work”. It answers questions, writes documents, makes images and media, and writes and runs code, all from one prompt box.
The agent runs in Google’s cloud, so a task can keep going for hours or days after the user closes the laptop. Google says users give it goals, not step-by-step instructions.
It can start helper agents. Some of them, which Google calls coworker agents, get their own email address and storage, and see only the information they are given. VentureBeat and TechCrunch report that each one gets its own Workspace account, calendar and Drive.
From Google Cloud’s announcement and reports by TechCrunch, VentureBeat and CNBC.
Gemini and Claude
The agent chooses a model for each task, and users can also pick one themselves, TechCrunch reports. Today it can use Google’s Gemini models and Anthropic’s Claude models. Google says other private and open models will follow.
It works inside Gmail, Docs, Sheets, Drive and Calendar. Users can also reach it from Microsoft 365, Slack, the command line, the web and phone apps. It connects to tools such as Salesforce, Jira and Snowflake, and to any MCP server.
To control costs, companies can set spending limits that pause the agent when it reaches them, Google says.
What is not known yet
Google gave no price and no date for when most customers can use the main agent, VentureBeat reports. Versions for finance and legal work are in preview, and versions for government, healthcare and retail are “coming soon”.
Google says the sportswear company On tested the agent early. No outlet has published an independent test, and how reliably the agent finishes long tasks is Google’s own claim.
Google also says that about 80% of its cloud customers use its AI products. These usage numbers are Google’s own, too.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2
- 3Research · October 8, 2026Google Cloud unveils persistent Gemini Agents for long-running tasks, and they get their own Gmail, Calendar, and Drive storage VentureBeat
- 4
- 5Research · October 8, 2026Google Cloud introduces Gemini agent to change enterprise work SiliconANGLE
What the agent does comes from Google’s announcement and reports from the event. How reliably it handles long tasks, and the usage numbers, are Google’s claims; no outside test exists yet.
Google launched one work agent that runs long tasks across many tools, though price, launch date and reliability remain unproven. How we rate the mood