Gemini Notebook gains official API and MCP server in public preview

The release covers only half of what Gemini Notebook does. Sources can now go in from code, but the steps that make the tool useful — asking questions of the material, generating explanations, and retrieving the results — are not in the API or the tool list. That split explains why the author frames this as a first step rather than a finished integration.

The practical argument for an API over a browser-driving agent is reliability and control: a predefined call is less likely to stall than an agent reading a screen, and the range of actions handed to AI can be narrowed. The author also noted that youtu.be short links come back normalized to the www.youtube.com/watch?v= form, a small sign the endpoints are already handling input consistently.

Adoption carries paperwork. The roadmap is a six-item list with no order or timing, and Google's note that it may change means it should be read as a plan, not a commitment. Until recall and generation join the API, Gemini Notebook remains something a person opens in a browser, with automation limited to filling it.