Google’s Gemini Agent Explained: One Universal AI Agent for Work, With Its Own Email and Drive
Google wants one agent to do it all. At Gemini at Work 2026, it introduced the Gemini agent: a universal work agent that chats, runs multi-day tasks, writes code, and can even get its own company email address. Here’s what we know.
What’s in this guide
What Google Announced
On October 8, 2026, Google Cloud CEO Thomas Kurian introduced the Gemini agent, described as “a single, universal agent for work.” Instead of separate assistants for chat, documents, data and coding, Google is folding them into one agent that can be chatted with, assigned objectives, scheduled, or triggered by events.
According to Futurum Group’s analysis, it runs on web, iOS, Android, Windows, Mac and the command line — and inside Google Workspace, Microsoft 365 and Slack.
Key Features
Multi-day tasks
Runs persistently in the cloud, so long jobs keep going after you close your laptop.
Sub-agents
Spins up temporary helper agents for parts of a bigger job.
Four kinds of memory
Session, semantic, procedural and episodic memory for context across days.
Skills
Reusable workflows — and the agent can write new skills for itself.
It connects to a tools registry that includes Salesforce, ServiceNow, Jira, Snowflake, Databricks, Microsoft Office and Teams, plus any MCP server.
Coworker Agents With Their Own Identity
The most striking idea is coworker agents: persistent agents that get their own @agents.company.com email address, calendar, Drive storage and Workspace account. Teammates can add them to Chat spaces or @mention them in Docs.
Crucially, these agents act under their own identity, not yours, and only see what people share with them. Each agent’s identity is cryptographically attested, and every action is logged in a per-agent audit trail. A Google representative told Futurum the company doesn’t plan to charge a per-seat license for coworker agents in Workspace.
Coworker agents act under their own identity — not yours — and only see what you share with them.
Multiple Models Under the Hood
The Gemini agent isn’t locked to Gemini. Its Smart Routing orchestrates across Google’s Gemini models and Anthropic’s Claude models today, with Gemma and other open and private models planned. That lets companies send simple, high-volume tasks to cheaper models and save top models for hard problems — without rebuilding workflows when they switch.
Security and Cost Controls
- Agent Sandbox: code runs in isolated environments.
- Agent Gateway: what Google calls an “AI network firewall” for enforcing governance policies.
- Permissions: OAuth-propagated access and action-level audit logs per agent.
- Spend caps: hard, real-time per-project caps in Cloud Billing that pause the agent when reached.
- Free re-runs: saved reports and skills can run on demand without new token costs.
Availability and Pricing
Google’s startup blog says the Gemini agent is not publicly available yet; interested companies can contact sales. Google hasn’t published pricing or general-availability dates for most announcements. Financial services and legal specializations are in preview, with government, healthcare and retail coming.
What It Means for Businesses
Google’s bet is that companies don’t want dozens of disconnected AI tools — they want one agent that works everywhere their people work, with strong identity and spending controls. If it delivers, the biggest change will be organizational: deciding which “coworker agents” to create, what they’re allowed to see, and who is accountable for their work.
FAQ
What is the Gemini agent?
A single, universal AI agent from Google Cloud that handles chat, knowledge work, data analysis, media creation and code, and can run multi-day tasks.
Can the Gemini agent use Claude?
Yes. Its Smart Routing works across Gemini and Anthropic Claude models today, with more models planned.
What are coworker agents?
Persistent agents with their own email address, calendar, Drive and Workspace account that act under their own identity.
Is it available now?
Not publicly yet. Google says interested businesses can contact sales; pricing hasn’t been announced.
Related Reading on FutureLume
- Gemini 4 Is Coming Early: Everything We Know So Far
- OpenAI’s Agents API Explained: What It Means for Teams Building AI Agents
- Computer-Use AI Agents in 2026: Which Tools Can Actually Control Your Screen
The Gemini agent is Google’s most ambitious answer yet to “which AI tool should we use?” — one agent, many models, its own identity. It isn’t broadly available yet, but it’s a clear preview of how AI coworkers will be managed in large companies.
Sources: Futurum Group · Google Cloud Blog · Neowin
