Three routes, and which one fits
NotebookLM, renamed Gemini Notebook in July 2026, now has two official APIs, both in preview, plus a set of community libraries. They serve different accounts and do different things.
| Gemini Notebook API | Gemini Notebook Enterprise API | Unofficial libraries (such as notebooklm-py) | |
|---|---|---|---|
| Who it is for | Personal Google accounts | Licensed users of Gemini Notebook Enterprise in Google Cloud | Any account that can log in to the web app |
| Status | Early access Public Preview | Preview (v1alpha), Pre-GA terms | Community project, no Google support |
| Notebooks | Create, list, update the name | Create, get, list recently viewed (up to 500), delete, share | Create, manage, share |
| Sources | Add web pages, YouTube videos, plain text, Markdown | Separate guide for data sources | URLs, YouTube, files, Drive |
| Ask a question | Not documented | Not on the notebooks page | Yes |
| Audio, video, reports | Not documented | Not on the notebooks page | Yes |
| Auth | Application Default Credentials with the notebook scope | A signed-in user account via gcloud, not a service account | Browser login or cookies |
| Cost | Free during the preview | Enterprise license | Free, MIT license |
The Gemini Notebook API
This is the API most people searching for “NotebookLM API” want. Google describes it as a REST interface to create and manage notebooks and sources. During the preview, the operations are: create, list and update notebooks, and add web pages, YouTube videos, plain text or Markdown as sources. Workspace accounts are not supported yet.
Limits from Google’s rate-limit page:
- Per project: 100 requests per minute for reads and 100 for writes.
- Per user: 20 per minute for adding a source, 60 per minute for each other method.
- Per source: up to 500,000 words, and inline text up to 1 MB.
- The account’s plan still applies: 50 sources per notebook on the free tier, up to 600 on Ultra.
A 429 error means you hit a limit; Google advises retrying with exponential backoff. The same API is also available as an MCP server for AI agents, see Can you connect NotebookLM to Claude or Cursor with MCP?
The Gemini Notebook Enterprise API
Companies on Gemini Notebook Enterprise get a separate API on the Discovery Engine endpoint (us, eu or global). It runs as the signed-in user, who needs a license and the Cloud NotebookLM User role. It can create, fetch, delete and share notebooks (as reader or writer), which is useful for setting up notebooks for a team or cleaning them up. Google marks it as Preview under the Pre-GA terms, available “as is” with limited support.
The unofficial libraries
Next to the official APIs, community developers built libraries on the internal endpoints of the web app. notebooklm-py is the best-known: a Python library, command line tool and MCP server that can chat with a notebook, generate Audio Overviews, slide decks and reports, and download them. Its own README is clear about the price of that reach: “Unofficial Library - Use at Your Own Risk”, it uses undocumented Google APIs that can change without notice, and heavy use may be throttled. It logs in through your browser or reuses browser cookies, and its optional master token is described as a powerful full-account credential, with the advice to use a dedicated account.
The common mistake: building a workflow on the query you cannot make
Teams often plan “our agent asks NotebookLM and uses the answer”. With the official APIs that is not documented today: you can fill and organize notebooks, not ask them questions. The only way to query is an unofficial library, which ties a business process to cookies of a personal Google account and to endpoints Google can change at any time. For a prototype that is fine. For a process a client depends on, it is a risk you should name.
What this means for a team
An API makes it easier to load the same sources into notebooks, but the knowledge still sits per notebook, per account, inside Google. Another AI tool cannot read what a notebook knows through the official API. If the goal is that Claude, ChatGPT or Cursor know what the team knows, you need a layer those tools can all reach, not a feed into one notebook. More on that in How to give AI agents your company context.
Where Bombos fits
Bombos is a plugin that gives every AI tool your team uses one shared memory that the tools can read and write through supported connections: team knowledge, rules it learns from corrections, and what already sits in mail, documents and repos, with a trail of what each answer rests on. It runs in the cloud or on your own server. One developer can try it first with their own agents, next to a notebook they already use.
Related questions
Can I use the Gemini Notebook API with a Workspace account?
Not yet. Google's developer page says the API is available for personal Google accounts during the preview, and that Google Workspace accounts are not yet supported.
Does the API cost anything?
Google says the Gemini Notebook API is free to use during the Public Preview. The Enterprise API needs a Gemini Notebook Enterprise license for the user.
Can I generate an Audio Overview through the API?
The official Gemini Notebook API documents notebooks and sources only. Unofficial libraries like notebooklm-py can generate audio and other outputs, but through undocumented endpoints.
Is notebooklm-py made by Google?
No. Its README calls it an unofficial community project, not affiliated with Google, that uses undocumented Google APIs that can change without notice.
