Google has opened an early-access public preview of the Gemini Notebook API, giving developers an official way to create notebooks and add research material from code. A companion connection lets AI assistants do the same. The preview’s biggest omission is also one of the product’s main attractions: developers cannot yet ask a notebook questions or generate its podcast-style Audio Overviews through it.
For people using Gemini Notebook, formerly NotebookLM, the immediate benefit is less manual collection. A script could gather a reading list, create a notebook and add the relevant sources. Notebook-generated answers would still need to be accessed through Google’s apps, rather than this API.
What an agent can do today
An API is an interface that lets software work with a service. Google’s Model Context Protocol, or MCP, server makes those operations available as tools for compatible AI assistants. Its documentation includes setup instructions for Claude Code and Cursor.
The official server exposes seven tools covering notebook management and source ingestion. Here is what those tools amount to in practice:
| Task | Available in this preview |
|---|---|
| Organise notebooks | Create, list and rename notebooks; retrieve their metadata. |
| Add research material | Import web pages, YouTube videos, plain text and Markdown. |
| Check an import | List sources and retrieve metadata, including processing status. |
| Ask questions or produce outputs | No source-grounded Q&A or artifact-generation tool. |
| Upload files or remove material | Binary uploads, Google Drive URLs and deletion remain on the roadmap. |
A developer following a software project could, for example, collect its release notes and public documentation in a notebook. Google’s source guide sets useful boundaries: web pages must be public HTTPS URLs without a login wall or paywall, and YouTube videos must be public or unlisted and have captions. Imports process asynchronously, so an integration needs to check whether each source becomes ready or fails.
“Get notebook” does not mean “ask notebook”
The distinction is easy to miss in an API reference. Google’s notebook resource contains information such as its name, title and creation time. Retrieving that resource does not ask the model to explain the contents. Likewise, retrieving a source reports its metadata and processing state rather than a generated answer.
Google lists retrieving information from a notebook and generating artifacts on its roadmap, alongside Workspace accounts, file uploads, Drive URLs and deletion. It supplies no delivery dates and says the roadmap can change. An end-to-end workflow that collects articles, asks for a briefing and automatically makes a podcast therefore goes beyond the current consumer API.
Free in preview, with account and setup limits
The preview supports personal Google accounts. Workspace accounts are not yet eligible. Google’s authentication guide requires a Google Cloud project and an OAuth sign-in on behalf of the user; it does not accept a simple API key. That account must also have signed in to Gemini Notebook at least once.
Google provides a command-line quickstart, but developers building an application still need to handle authentication properly. The access tokens used in the documented local MCP setup expire after an hour. This is a developer integration with setup work, rather than a one-click connection for every Notebook user.
The API is free during public preview, subject to rate and capacity limits. The free tier allows 100 notebooks and 50 sources per notebook. Each user can create up to 20 sources a minute; project-level limits also apply. Higher subscription tiers allow more notebooks or sources, but Google has not made a promise here about API pricing after preview.
Three routes that should not be confused
Google already offers a separate Gemini Notebook Enterprise API that requires an active Enterprise licence. Calling this preview the first official Notebook API without the personal-account qualification would obscure that existing service.
There is also a Gemini Notebook Connected App inside Gemini, which Google announced on 23 October. That integration can answer questions using notebook content within Gemini chats. Its capabilities do not establish what an outside developer can call through the new API or MCP server.
For developers whose immediate problem is feeding sources into a personal research notebook, this preview covers a useful part of the job. Anyone building an application around cited answers or automatically generated podcasts still needs to account for the missing output operations.
Reporting note: This article describes Google’s published documentation, checked on 8 November 2026. We have not run a hands-on API test. The overview was last updated on 2 November; that date is not confirmation of a launch date. TLDR AI covered the preview on 4 November.
Discover more from Sulat AI & Tech News
Subscribe to get the latest posts sent to your email.
