The Google integration is two provider accounts, not one. Both let AI steps run on your own Google account; they reach different products, with different keys, quotas, and terms.
What it gives you
No Google nodes—those are Coming Soon.
What the accounts do today is serve models. Point the built-in AI Agent action or the AI Judgment condition at one, on the step, on the agent it runs as, or as your default model, and that work runs on your Google account: the tokens land on your bill, and your content travels only to the account you named.
What it needs
One of two connections, depending on which product you have access to:
| Connection | What it reaches |
|---|---|
| Gemini | The Gemini Developer API, with a key from Google AI Studio. One field to fill in. |
| Google Cloud | Gemini through Vertex AI, in a Google Cloud project you name, drawing on that project's quota. |
They are not two spellings of the same thing, and a key from one does not work on the other.