What Is the Gemini Agent?
Announced on October 8, 2026 at the Gemini at Work 2026 event, the Gemini agent is what Google Cloud CEO Thomas Kurian calls "a single, universal agent for work." Instead of juggling separate chatbots for writing, coding, and analysis, you give the agent objectives — not step-by-step instructions — and it answers questions, handles knowledge work, creates images and media, and writes and runs code from one interface and one API. Google says the agent holds your organization's business context, plans the work, connects to company systems, and returns finished work inside the documents, inboxes, and developer environments people already use.
Works Inline Where You Already Work
The agent lives directly inside Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar, carrying the same memory, skills, and controls everywhere. It can also be reached from any device — web, mobile, desktop — and through command line, Microsoft 365, and Slack, or it can run headless inside third-party applications. One real example from Google's announcement: when a manager emails asking for a project update as a slide deck, Workspace Intelligence recognizes it as a delegatable task and offers a one-click option to hand it to the agent.
- One interface and one API for questions, writing, media, and code
- Inline in Gmail, Drive, Docs, Slides, Sheets, Chat, and Calendar
- Reachable via web, mobile, desktop, command line, Microsoft 365, and Slack
- Connects to business systems like Salesforce, ServiceNow, BigQuery, and any MCP server
Smart Routing: The Right Model for Each Job
A standout feature is multi-model orchestration. The agent "runs each job on the model that fits best," routing tasks across the Gemini family and Anthropic's Claude models today, with more private and open models planned. This is meant to deliver optimal quality at lower cost — Google pairs it with real-time spend caps, so companies can pause work automatically when usage hits a limit. Early testers named by Google include PayPal, which routes 10 million multi-model requests per week, and Brazilian bank Bradesco, which cut document review time from one hour to five minutes.
Coworker Agents and Sub-Agents
For longer or team-scale work, the agent can spin up temporary sub-agents — each with its own identity — that coordinate multi-step workflows running for hours or even days. Going further, "coworker agents" act as full team members: they get their own Workspace account, an @agents.company.com email address, calendar, Drive, and directory entry, and only see information explicitly shared with them. Google also previewed industry-specific agents: financial services and legal versions are already in preview, with government, healthcare, and retail versions coming soon.
“Give it objectives, not instructions.”
| Capability | What it does |
|---|---|
| Smart Routing | Picks the best model per task across Gemini and Claude families. |
| Coworker agents | Team-member agents with their own Workspace account and email. |
| Inline in Workspace | Works inside Gmail, Docs, Sheets, Drive, and Calendar. |
| Industry versions | Financial services and legal in preview; government, healthcare, retail coming. |
Frequently Asked Questions
When will the Gemini agent be available?
Google announced it on October 8, 2026 at Gemini at Work 2026. It is rolling out to enterprise customers first, with industry-specific versions for financial services and legal already in preview and wider Workspace availability to follow.
Which AI models power the Gemini agent?
Today it routes work across Google's Gemini model family and Anthropic's Claude models, automatically picking the best model for each job. Google says support for additional private and open models is coming.
What is a coworker agent?
A coworker agent is an agent instance that acts as a team member with its own Workspace account, email address at @agents.company.com, calendar, Drive, and directory entry. It only sees information shared with it, so companies can scope its access carefully.
How does the Gemini agent control costs?
It uses multi-model orchestration and Smart Routing to run each job on the most cost-effective model that fits, plus real-time spend caps that pause work when usage hits a company-set limit. Costs can be tracked per project.
