Gemini Agent: What It Is, Features, How to Use It and UK/EU Availability (2026 Guide)

Google’s new Gemini agent is the company’s biggest bet yet on AI that does work rather than just answering questions. Announced by Google Cloud on 8 October 2026 at its Gemini at Work event, it is described as “a single, universal agent for work”: you give it an objective, it plans the steps, uses your tools and comes back with finished work, even if the job takes hours or days. This guide explains what the Gemini agent is, what it can do, how you will use it inside Gmail, Docs and other apps, what is known about pricing and availability in the UK and Europe, and how to prepare your organisation.

Everything below was checked on 9 October 2026 against Google Cloud’s keynote post, Google’s own announcement and reporting from 9to5Google, TechCrunch and VentureBeat. The Gemini agent is in an early preview, so details such as pricing and exact menus have not been published yet. Where Google has not confirmed something, we say so.

What Is the Gemini Agent?

The Gemini agent is an AI agent that lives behind a single prompt box and works across your company’s apps and data. Google Cloud CEO Thomas Kurian summed up the idea: “You give it objectives, not instructions. You delegate an outcome and come back to finished work.” It can answer questions, handle knowledge work, create images and media, and write and run code.

Google also stresses that the agent and the model are separate: “Gemini is the agent, and the model underneath it is a separate choice.” The agent routes each job to a suitable model, which today includes Google’s Gemini family and Anthropic’s Claude models, with other private and open models planned.

Is this the same as Gemini “agent mode”?

Not quite. Search suggestions show people also look for “Gemini agent mode”, which usually refers to agent features in coding tools such as VS Code or Android Studio. The Gemini agent announced on 8 October is a broader work agent for organisations. If you are an individual Gemini app user, TechCrunch reports Google will bring it to consumers later, but no date has been given.

Gemini Agent Features Explained

1. Objectives instead of step-by-step prompts

Rather than asking for one email or one summary at a time, you describe the result you want. Google’s example: ask Gemini to set up a meeting with “the usual team of regional event leads next week” without giving a single name. The agent works out who those people are from past chats and email threads, checks their calendars and emails them to agree a time, including people outside the company.

2. Long-running work in the cloud

The agent runs in Google’s cloud, so “work that takes hours or days keeps running after you close your laptop.” For bigger jobs it can create temporary, job-specific sub-agents and coordinate steps in parallel or in sequence. TechCrunch reports a “tasks inbox” that shows the agent’s reasoning, which sub-agents it used, which skills it loaded and its progress.

3. Four kinds of memory

Google says the agent keeps four types of memory, which is how it gets better the longer you work with it:

• Session memory: the current task, even across a multi-day run.
• Semantic memory: a structured knowledge base built from documents, conversations and work with other agents.
• Procedural memory: how jobs get done, including skills the agent writes for itself.
• Episodic memory: a record of everything it has done before.

4. Coworker agents with their own identity

This is the most unusual part. Besides acting as your personal assistant, Gemini can create “coworker agents” for a role, such as a project manager or a finance analyst. Google says a coworker agent gets its own Workspace account, including an email address (for example on an @agents.company.com domain), calendar, Drive storage and a place in the company directory. It acts under its own name, its edits show up under that name in version history, and it only sees what people share with it. To create one, Google says you simply describe the role you need.

5. Tools, skills and connectors

Google lists connections to Confluence, Microsoft Office and Teams, Slack, Workspace, Git, Jira, Salesforce, ServiceNow, BigQuery, Databricks, Postgres, Snowflake, local desktop files and any Model Context Protocol (MCP) server. Teams can also publish reusable skills to a shared skills registry, an idea similar to the personal skills we covered in our guide to Gemini Skills.

How to Use the Gemini Agent

Because the agent is in preview, Google has not published a full help guide yet. Based on the workflows Google demonstrated, this is how using it is expected to work once your organisation has access:

1. Check you are eligible. Ask your Google Workspace or Gemini Enterprise administrator whether your organisation is in the preview or on one of the Business or Enterprise plans Google says will get it first.
2. Start from the app you are already in. Google says you can @mention Gemini in Gmail, Docs, Sheets, Slides and Chat, or tag the agent in a document comment.
3. Delegate with one click. When Workspace spots a request that can be handed off, such as a manager asking for a project update as a slide deck, it offers a single-click option to pass the task to Gemini.
4. Describe the outcome, not the steps. Say what “done” looks like, who it is for, the deadline and any sources it should use.
5. Follow progress. Check the tasks inbox to see the plan, the sub-agents involved and what has been completed, and step in if it heads in the wrong direction.
6. Create a coworker agent for repeat work. Describe a role, such as “events coordinator for the marketing team”, share only the files and spaces it needs, and add it to the relevant group chat.
7. Use it outside Google’s apps. Google says the same agent is reachable from Slack, Microsoft 365, the command line and mobile and desktop apps, and developers can embed it in other software as a “headless” agent through an API.

Example objectives to try

• “Turn last month’s customer feedback emails into a five-slide summary with the top three issues and suggested fixes.”
• “Book a 30-minute review with the Dublin and Berlin sales leads next week and send them the latest pipeline sheet beforehand.”
• “Compare our Q3 marketing spend in BigQuery with the budget in Sheets and flag anything over 10%.”
• “Draft a launch-readiness document from the project chat and post it back to the group when it is done.”

If your team already keeps research in notebooks, our Gemini Notebook (NotebookLM) guide explains how source-grounded notebooks work, which is useful background for giving an agent reliable material.

Security and Governance

An agent that can send emails and edit files needs firm limits. Google frames governance around four questions: who the agent is, what it may do, what it did, and what it must never touch. The controls it describes are:

• Identity: every agent gets its own cryptographically attested identity, “governed like an employee” with least-privilege permissions.
• Authorisation: role-based permissions approved by your security administrators, with external identities passed through standards such as OAuth.
• Audit trail: every action is logged and attributed to the agent rather than to a person.
• Agent Sandbox: tasks run in a sandbox with its own network boundary.
• Agent Gateway: all traffic in, out and between agents passes through what Google calls an AI network firewall that enforces company policies in real time. Google’s example policy: agents may not open documents classified “Need to Know”.

Sandboxing agents is becoming standard practice across the industry; our guide to NVIDIA OpenShell shows how an open-source sandbox approaches the same problem.

Gemini Agent Pricing and Cost Controls

Google has not announced a price for the Gemini agent, and VentureBeat notes it is unclear whether it will be included in existing Gemini Enterprise subscriptions. What Google has described are cost controls:

• Multi-model orchestration and smart routing: each job is sent to the model that gives the required quality at the lowest cost.
• Real-time spend caps: administrators can set a hard limit per project in the Cloud Billing console. If the cap is reached, that project’s agent pauses and can be resumed with one click.
• Cost attribution: spending is tracked per project, so it can be charged back to departments.

Until official pricing appears, treat any figures you see online with caution and budget a pilot with a strict spend cap.

Gemini Agent in the UK and Europe

Google has not published country-by-country availability for the Gemini agent. For organisations in the UK, Ireland, Germany, France, the Netherlands, Spain, Italy, Sweden and Poland, a few practical points apply:

• Data location: Gemini Enterprise already offers an EU processing location for some models, according to its release notes. Ask Google which locations the agent will support before rolling it out.
• GDPR: an agent with its own mailbox and calendar will process personal data. Involve your data protection officer early and update records of processing.
• Works councils: in countries such as Germany and the Netherlands, introducing tools that log employee activity may require consultation.
• Multi-model routing: if the agent sends some work to Claude models, check how that affects your data processing agreements and where data is handled.

How to Prepare Your Organisation

1. Pick two or three repetitive workflows with clear outcomes, such as meeting scheduling, weekly reports or first-draft documents.
2. Tidy permissions. Agents follow your existing sharing rules, so over-shared folders become an agent’s problem too.
3. Write policies once. Decide what agents must never access and how sensitive documents are labelled.
4. Set a budget and a spend cap for the pilot.
5. Define human checkpoints for anything sent outside the company, such as client emails.
6. Train staff to write objectives. Clear goals, audiences and deadlines matter more than clever prompts.

Gemini Agent vs Other AI Agents

The Gemini agent arrives in a crowded field. Microsoft has rebuilt Copilot around agents, as we explain in our report on Copilot’s new Autopilot agents, and consumer agents such as Meta Muse target personal tasks. People are also searching for comparisons with Anthropic’s Claude Cowork.

Google’s pitch rests on three differences: the agent lives inside Workspace where many companies already work, coworker agents get their own accounts and audit trails, and the agent is model-agnostic, able to use Claude as well as Gemini. Google already offers enterprise voice agents too, as covered in our article on Gemini Live Avatar.

Pros and Cons

Pros

• Works inside Gmail, Docs, Sheets, Slides, Chat and Calendar, plus Slack and Microsoft 365
• Handles long, multi-step jobs that keep running in the cloud
• Coworker agents with their own identity, permissions and audit trail
• Can choose between Gemini and Claude models
• Built-in spend caps and smart routing

Cons

• Private preview only; most organisations cannot use it yet
• No published pricing or general availability date
• Consumers must wait, with no date announced
• Regional availability and data residency not yet detailed
• Customer results so far are Google’s own claims

Frequently Asked Questions

What is the Gemini agent?

It is Google’s “universal agent for work”, announced on 8 October 2026. You give it an objective, and it plans the work, uses tools and connected apps, and returns finished results.

Is the Gemini agent free?

Google has not announced pricing. It is aimed at businesses first, through Google Workspace and Gemini Enterprise, and it is unclear whether it will be included in existing subscriptions.

How do I get access to the Gemini agent?

It is in private preview. Google says wider availability is coming soon for Workspace customers on select Business and Enterprise plans, so ask your administrator or Google account team.

Is the Gemini agent available in the UK and EU?

Google has not published regional availability. Check with Google before planning a rollout, especially if you need EU data processing.

Does the Gemini agent use Claude?

Yes. Google says the agent can route work to Gemini models and to Anthropic’s Claude models today, with other models planned.

Can consumers use the Gemini agent?

Not yet. According to TechCrunch, Google will roll it out to businesses first and to consumers later, without giving a date.

Conclusion

The Gemini agent marks a clear shift from chatbots to delegated work: one agent, inside the apps your team already uses, able to plan long jobs, create coworker agents and pick the right model for each task. For now it is a preview for businesses, with no price and no consumer date. The smart move is to prepare: choose a few repeatable workflows, tidy your permissions, plan for GDPR and set a strict budget, so you are ready to test it as soon as Google opens access.

Sources: Google Cloud – Welcome to Gemini at Work 2026; Google – Google Cloud introduces the Gemini agent; 9to5Google; TechCrunch; VentureBeat; Gemini Enterprise release notes.