AI Explained
Google Gemini, product by product
One family of models, three ways in: at work, in the terminal, and on the platform where teams build their own agents.
The map
One family of models, three ways in
Gemini is the name of Google's models and of the products built on them. Play the diagram to see which door is for which job.
Press Play to walk through it, or Listen to have it read to you.
Read all the steps
- Gemini is one family of models with three ways in. Which one you use depends on the job in front of you.
- At work: Gemini inside Gmail, Docs, Drive and Meet answers using your own Workspace content. Google says that content is not used to train the models.
- In code: Gemini CLI is an agent in your terminal. GEMINI.md files give it your project's context, and MCP servers, extensions and skills give it tools.
- Building: the Agent Platform, the evolution of Vertex AI, is where technical teams code, scale and govern their own agents.
- Behind every door is a Gemini model. What changes is what the model is given: your documents, your repository, or your own agent's tools.
Start using it
At work, in the terminal, in the editor
Eight steps, with the commands from the documentation. Copy a command, paste it in your terminal.
At work: ask inside the app
In Gmail, Docs, Drive, Sheets, Slides or Meet, Gemini uses your Workspace content to answer, and Google says it does not use that content to train its models [1].
In code: install Gemini CLI
It installs with npm [2].
npm install -g @google/gemini-cliStart it and sign in
Run it, choose Sign in with Google and pick your account. Some account types need a Google Cloud project [2].
geminiTry a first task
The quickstart suggests asking it to explain how a repository works from its source, or to write unit tests for a file. Approve the file changes when it asks [2].
Give it your project's context
Put your rules and facts in GEMINI.md context files, which are read as a hierarchy [3].
Add tools, keep it safe
Connect MCP servers, extensions and skills. Use the read-only plan mode for a large change, and sandboxing to isolate what its tools run [3].
Check your usage
The quickstart gives this command for token usage and quota [2].
/stats modelIn the editor
Gemini Code Assist has an agent mode in VS Code and IntelliJ that works with built-in tools and MCP servers [4].
Which one?
Pick the right door
Choose what you are trying to do. The guidance is this center's own, built from the descriptions below.
The family
Product by product
As Google's pages described them on the date in the sources. Google renames these products often: follow the links for the current names and plans.
Gemini in Google Workspace
Gemini inside Gmail, Docs, Drive, Sheets, Slides and Meet uses your Workspace content to give more useful answers, and Google says it does not use that content to train or improve its generative models [1].
Gemini Enterprise
Google Cloud calls it its agentic platform. The Gemini Enterprise app is a subscription service, billed for each seat, where business users run and design no-code agents [5].
Gemini Enterprise Agent Platform
The developer side. Google describes it as the evolution of Vertex AI, built for technical teams to code, scale and govern agents, and billed by consumption on a Google Cloud account [5].
Gemini CLI
It brings Gemini models into your terminal [3]. You install it with npm, run it and sign in with Google [2].
GEMINI.md and memory
Project context lives in GEMINI.md files, read as a hierarchy, and a memory feature manages persistent instructions and facts [3].
MCP servers and extensions
MCP servers bring in outside tools, and extensions add new tools and capabilities [3].
Skills, subagents and hooks
Skills and subagents bring specialized help for particular tasks, and hooks customize its behaviour with your own scripts [3].
Sandboxing, trusted folders and plan mode
Sandboxing isolates what its tools run, trusted folders are part of its permission checks, and plan mode is read-only, for planning complex changes [3].
Checkpointing and headless mode
Checkpointing takes automatic snapshots of a session, and headless mode is the interface for scripts [3].
Gemini Code Assist, agent mode
Available in VS Code and IntelliJ. It uses tools, which can be built-in system tools or MCP servers, to get context and to act [4].
Check which CLI you get
The Gemini CLI documentation carries a notice that for unpaid tier and Google One users it was replaced by Antigravity CLI on June 18, 2026 [2]. Read the notice before you install.
This center is independent and is not affiliated with Google. The product names belong to their owner.
Before you rely on it
Three habits
Know what it can read
An assistant answers from what its account, or your repository, lets it open. Tidy the sharing and the secrets before a wide rollout, not after.
Treat what it reads as untrusted
A document, a web page or a tool result can carry text written to steer the model. Limit what the tool may do with what it reads, and test for it.
Keep a person on what cannot be undone
Let it draft, plan and propose freely. Sending, deleting, paying and deploying wait for a person, and every action is logged.
Sources
Where this comes from
Read 2026-11-15, unless a source gives its own date. A number in brackets points to the page a statement was read on. Sentences with no number, the small demonstrations and the tools are this center's own explanations.
- Google Workspace Help: how Gemini in Gmail, Docs, Drive and Meet protects your data Read 2026-11-16.
- Gemini CLI: Get started Read 2026-11-16.
- Gemini CLI documentation Read 2026-11-16.
- Gemini Code Assist: Agent mode overview Read 2026-11-16.
- Google Cloud: Gemini Enterprise frequently asked questions Read 2026-11-16.