# Use Gemini Notebook (formerly NotebookLM) Studio in Claude or ChatGPT
> Use NotebookLM Studio from Claude or ChatGPT: start Audio Overviews in 4 formats, slide decks, quizzes and flashcards, then check, fetch and export them.
By Juli · Published Oct 1, 2026 · Updated Oct 2, 2026 · https://www.kortex-notebooklm.com/blog/gemini-notebook-studio-from-chat
You're deep in a Claude chat. The notebook is ready, and you want a 10-minute Audio Overview for the drive home. So you open another tab, find the notebook, click Studio, pick a format, retype your steering prompt and wait. Then you do it all again for the deck. That tab shuffle is exactly what 91 of the 776 people on our Kortex MCP waitlist asked us to remove (Kortex MCP waitlist survey, 776 respondents, September 19 to October 1, 2026). Good news: it already works. With Kortex MCP, Claude or ChatGPT can start any Gemini Notebook (formerly NotebookLM) Studio output, steer it with your instructions or a saved prompt, check when it's done, and hand back a file, a Google Doc or a podcast episode. Here's the whole loop, with prompts you can paste.
> **Key Takeaways**
> - 91 of 776 Kortex MCP waitlist respondents asked to make audio, slides or quizzes from chat (Kortex MCP waitlist survey, 2026). It works today.
> - One tool makes Audio Overviews in 4 formats; another covers 10 more Studio types.
> - Steer outputs with plain instructions or a saved Kortex prompt.
> - Quiz answers come back into the chat, though Gemini Notebook has no quiz download.
> - Kortex MCP is a Pro feature in private beta, by invite.
## What can Claude or ChatGPT make in Gemini Notebook Studio?
Everything in Studio. In short, **Studio** is the Gemini Notebook panel that turns your sources into audio, decks and study material. Kortex MCP starts an Audio Overview plus 10 other Studio types from chat, inside your own Gemini Notebook account (Kortex MCP tools docs, 2026). You ask in plain words, and the assistant picks the tool and fills in the options.
| Output | Tool | What you can set from chat |
|---|---|---|
| Audio Overview | `generate_audio_overview` | Deep dive, brief, critique or debate; short, default or long. |
| Slide deck | `generate_artifact`, type `slides` | Detailed deck or presenter slides; short or default length. |
| Quiz, flashcards | `generate_artifact`, type `quiz` or `flashcards` | How many (short, medium, long) and difficulty (easy, medium, hard). |
| Mind map | `generate_artifact`, type `mindmap` | Instructions and which sources to use. |
| Reports | `generate_artifact`, type `briefing_doc`, `study_guide` or `blog_post` | Add to Google's built-in brief, or replace it. |
| Video Overview | `generate_artifact`, type `video_overview` | Explainer, brief, cinematic or vertical short; a style like whiteboard or watercolor. |
| Infographic | `generate_artifact`, type `infographic` | Landscape, portrait or square; detail level; a style like sketch note. |
| Data table | `generate_artifact`, type `data_table` | Instructions and which sources to use. |
Every type also takes a language code and a list of sources to stick to. For example, "make Spanish flashcards from chapters 3 and 4 only" is one request, not three menus.
One thing doesn't change: Gemini Notebook still does the making. In fact, Kortex has no text-to-speech of its own (Kortex docs, 2026). It asks Studio from your signed-in Chrome, just like a click; as a result, what you get matches the tab.
## Why can't Claude or ChatGPT use Studio on their own?
Because there's no door for your assistant to walk through. Google publishes no public consumer API for Gemini Notebook. Its only documented API is enterprise-only, through Gemini Enterprise on Google Cloud ([Google Cloud](https://docs.cloud.google.com/gemini/enterprise/notebooklm-enterprise/docs/api-notebooks), 2026). Therefore, Claude and ChatGPT can't reach Studio by themselves. Two smaller gaps show up once you try. First, quizzes have no download at all: Google's help offers flashcards as a CSV file and lists nothing for quizzes ([Gemini Notebook Help](https://support.google.com/gemininotebook/answer/16958963?hl=en), 2026). Second, the steering prompts you've polished live in your head or a notes app, so you retype them every time.
Hence the job runs through you. We built Kortex MCP, and when we analyzed the 776 waitlist answers, we found 244 people (31%) describing a manual copy-paste workaround they use today (Kortex MCP waitlist survey, 2026). Making Studio outputs from chat tied for fourth among everything they wanted their assistant to take over, with 91 people. Only scheduling (128), tidying (121) and cited answers (110) ranked higher. Notably, 43% of those 91 already copy outputs by hand, and 41% pay for Kortex Pro. For example, one wrote that they wanted to "generate template-based audio summaries, video summaries, written summaries, infographics." In other words, they want the Studio they already use, started from the chat already open.
Source: Kortex MCP waitlist survey, 776 respondents, September 19 to October 1, 2026.
For Studio work, the workaround goes like this. Open the notebook, retype the prompt and wait. Then paste the result into the chat, or alternatively re-upload the file. Every output costs you that trip. However, the fix starts with a free Chrome extension.
Skip the tab shuffle, starting with Kortex free
Get the free Kortex extension for Gemini Notebook today. Making Studio outputs from chat runs through it with Kortex MCP, a private beta for invited Kortex Pro members, so join the waitlist for an invite.
## How does Kortex MCP connect your chat to Studio?
**Kortex MCP** is a bridge between your AI chat and Gemini Notebook. Your AI client calls `https://mcp.kortex-notebooklm.com/mcp`, the call is relayed to your own Kortex Chrome extension, and the extension talks to Gemini Notebook with the Google session you're already signed into (Kortex MCP docs, 2026). Kortex never stores your Google password or cookies. The bridge runs both ways. Whatever your AI client can reach, Kortex can put into a notebook with `add_text_source` or `add_url_source`. That could be a draft you pasted, a page it read, or a file it opened with its own connectors, such as Claude's Google Drive connector ([Claude Help Center](https://support.claude.com/en/articles/10166901-use-google-workspace-connectors), 2026). Studio then turns that material into audio or slides, and Kortex carries the result back out to you. Kortex itself never reads your Gmail or Drive; the assistant does the fetching. Our [Gemini Notebook MCP guide](/blog/notebooklm-mcp) compares other routes.
We use Claude and ChatGPT here because our waitlist named them most: Claude 126 times, ChatGPT 69 (Kortex MCP waitlist survey, 2026). Likewise, the same tools work from Cursor, Codex CLI and other MCP clients.
### Before you start
Kortex MCP is a Kortex Pro feature, and only invited beta members can connect. Once you have an invite, however, connecting takes a minute:
- **Claude (web or desktop):** add a custom connector with the endpoint URL, then sign in on the Kortex consent page ([Claude Help Center](https://support.claude.com/en/articles/11175166-get-started-with-custom-connectors-using-remote-mcp), 2026).
- **ChatGPT:** turn on developer mode under Settings, then Security and login, and add Kortex as an MCP app with OAuth. Developer mode is on Plus, Pro, Business, Enterprise and Education plans, on the web ([OpenAI](https://developers.openai.com/api/docs/guides/developer-mode), 2026).
- **Claude Code:** one command.
```bash
claude mcp add --transport http kortex https://mcp.kortex-notebooklm.com/mcp
```
Keep Chrome running with the extension signed in; notably, no notebook tab needs to be open. That said, the first call after Chrome has been idle can take up to 30 seconds while the extension wakes up. For the full walkthrough, see [using Gemini Notebook in Claude](/blog/gemini-notebook-in-claude).
## How do you make an Audio Overview from Claude or ChatGPT?
Ask for it, and name the format and length. `generate_audio_overview` takes 4 formats (deep dive, brief, critique, debate) and 3 lengths (short, default, long) (Kortex MCP tools, 2026). However, brief has a single length, and long works only with deep dive. Meanwhile, the tool returns an id right away while Gemini Notebook records.
```
In my "Climate Policy" notebook, make a debate-format Audio Overview,
default length. Have the hosts argue carbon tax vs cap-and-trade for
a listener who already knows the basics. Use only the three 2026
reports, not the older sources. Tell me when it's ready.
```
That last line matters: the assistant can't hear the audio, but it can keep checking until the episode is finished.
Want it in another language? Say so. Audio Overviews come in over 80 languages; however, Google notes the length control is English-only ([Gemini Notebook Help](https://support.google.com/notebooklm/answer/16212820?hl=en), 2026). Not sure which format fits? Our [guide to Gemini Notebook podcasts](/blog/notebooklm-podcast) compares all four.
### Know when it's ready
Ask, or tell the assistant to keep checking; this works for every Studio output. `list_artifacts` reports each item as in progress, completed, failed or unknown, with an error reason for failures and a media link for finished audio or video (Kortex MCP tools, 2026). Generation usually takes one to several minutes.
Leave out the notebook and it sweeps your whole library in one call, including jobs you started in an earlier chat.
```
Check every notebook for Studio items I started today. List what's
finished, what's still running and anything that failed, with the
reason. If something is still running, check again in a few
minutes. Don't regenerate anything.
```
That last sentence saves quota. If a generation runs past the relay's time budget, the tool hands back a job id, and the assistant should poll `get_job`. In contrast, calling the generate tool again starts a second run, which spends your Gemini Notebook usage twice.
Since September 2, 2026, that usage refreshes every 5 hours until a weekly limit ([Gemini Notebook Help](https://support.google.com/gemininotebook/answer/17670842?hl=en), 2026). Our [usage limits guide](/blog/gemini-notebook-usage-limits) explains what uses more of it. Accordingly, if a run fails on a limit, `get_notebooklm_usage` shows which Google-side limit your account hit.
## How do you steer Studio outputs with your own prompts?
Type instructions in the chat, or reuse a prompt from your Kortex library. Both generate tools accept free-text `instructions` or a saved `prompt_id`. Notably, if you give both, the saved prompt becomes the body and your instructions refine it (Kortex MCP tools, 2026). Consequently, one good prompt serves every notebook.
### Plain instructions
```
Make a presenter-slides deck from my "Q4 Board Prep" notebook, short
length. Audience: the board, not the product team. One idea per
slide, lead with revenue and churn, end with three decisions we
need from them.
```
Reports behave a little differently. Specifically, for briefing docs, study guides and blog posts, your instructions are added to Google's built-in brief by default. Ask the assistant to replace the preset when you want your words to be the whole brief.
### Saved prompts with blanks
Saved prompts can hold `{{placeholders}}` that the assistant fills on each run, and anything left unfilled comes back flagged. You can save one without leaving the chat:
```
Save this to my Kortex prompts as "Exam quiz":
Write questions only on {{topic}}. Mix recall and application. Make
every wrong option a mistake a {{level}} student really makes.
Then make a hard, long quiz from my "Bio 201" notebook with it.
Topic: cellular respiration. Level: first-year.
```
This is the quiet win. In other words, the prompt that took you six tries to get right becomes a tool your assistant can call by name, with new blanks each time.
## How do you get quiz, flashcard and mind-map content into the chat?
Use `get_artifact_content`. It reads the cards in a flashcard deck, each quiz question with its correct option, or the tree of a mind map (Kortex MCP tools, 2026). It matters most for quizzes, which Gemini Notebook itself won't let you download.
You get structured JSON by default, or one Markdown document for flashcards and quizzes (mind maps come as a tree only). A panel also opens in the chat for working through the deck or quiz. Your assistant can then:
- Quiz you one question at a time and grade your answers.
- Rewrite a flashcard deck as lines you can import into Anki.
- Turn a mind map's tree into an indented outline for your notes.
```
Get the content of the newest quiz in my "Bio 201" notebook. Ask me
the questions one at a time, wait for my answer, then tell me if I'm
right using the quiz's own answer key. Keep score.
```
Notice the phrase "the quiz's own answer key". The correct option travels with each question; as a result, the assistant grades you against the quiz Gemini Notebook wrote from your sources. It isn't guessing from its own memory of biology, which is where chat-made quizzes tend to drift. Pair this prompt with the "Exam quiz" one above and that's the whole loop in one chat. Why bother being quizzed at all? Testing yourself beats rereading. In a classic 2006 study, students who took recall tests remembered much more a week later than students who restudied the same passages, even though the rereaders felt more confident ([Roediger and Karpicke, PubMed](https://pubmed.ncbi.nlm.nih.gov/16507066/), 2006). For a full exam routine, see [how to study with Gemini Notebook](/blog/notebooklm-for-studying).
Your own quiz, asked and graded right in the chat
Kortex MCP lets Claude or ChatGPT make the quiz, deck or Audio Overview and bring it back into the conversation. It's a private beta for invited Kortex Pro members, so join the waitlist for an invite.
## Can you revise a slide deck from chat?
Yes, slide by slide. `revise_slide_deck` takes a list of slides, counted from zero, and an instruction for each (Kortex MCP tools, 2026). Importantly, until you confirm, a call changes nothing. Instead, it returns which slides it would touch, and the revision runs only after your yes.
```
Revise my latest "Q4 Board Prep" deck. Slide 2: cut the text to
three bullets. Slide 5: replace the chart with one big number.
Slide 9: make the ask a single sentence. Show me the changes before
you run them, and tell me if the original deck survived.
```
Two facts shape how you revise. First, the revision should create a new deck and leave the original alone, and the reply says whether it did. In addition, Google says "sources are not taken into account during revisions" ([Gemini Notebook Help](https://support.google.com/gemininotebook/answer/16757456?hl=en), 2026). Thus you should revise wording, layout and visuals. For wrong facts, fix the sources and generate a fresh deck.
## How do you get Studio outputs out of Gemini Notebook?
Three ways. First, `export_artifact` saves a file, a Google Doc or a Google Sheet. Second, `get_formatted_content` pulls the text into the chat. Finally, a Kortex podcast sends finished Audio Overviews to your podcast app (Kortex MCP tools, 2026).
### Export to a file, Docs or Sheets
`export_artifact` has 3 destinations. `download` returns a media link or inline Markdown or CSV. In contrast, `google_docs` and `google_sheets` use Gemini Notebook's own Drive export, and you can name the Drive file in the same request. Similarly, Google's help says data tables export to Sheets with each table on its own tab ([Gemini Notebook Help](https://support.google.com/gemininotebook/answer/16262519?hl=en), 2026).
```
Export the study guide from my "Bio 201" notebook to Google Docs,
titled "Bio 201 Unit 3 study guide". Then export the data table from
the same notebook to Google Sheets and give me both links.
```
Want the text in the chat instead? `get_formatted_content` returns notes and artifacts as Markdown or plain text, with no download and no Kortex export credits spent. Consequently, your assistant can carry the text anywhere its own tools reach. Our guide to [moving notebook outputs into Notion or Obsidian](/blog/gemini-notebook-to-notion-obsidian) covers that half.
### Put Audio Overviews in your podcast app
Add them to a Kortex podcast. A **Kortex podcast** is a private RSS feed you follow in Spotify, Apple Podcasts or any app that adds shows by URL (Kortex docs, 2026). `create_podcast` makes the show, `add_podcast_episode` adds a finished Audio Overview, and `get_podcast_feed_url` returns the link.
```
When the climate debate episode is finished, add it to my podcast
"Commute Listening". If that podcast doesn't exist, create it first.
Then give me the RSS feed URL so I can add it in Overcast.
```
Three details worth knowing before you hit play:
- **Open the dashboard once.** The RSS audio file is prepared the next time you open the Kortex dashboard.
- **Minutes are a reusable pool.** Each episode is charged once, rounded up to the next minute. Pro's 1,000 minutes are a one-time balance, and removing an episode refunds it.
- **Treat the feed link as private.** Anyone with the link can listen, and the feed includes your account email in the owner field.
Want an episode every Monday without asking? For instance, a Kortex schedule can generate an Audio Overview and add it to your podcast while Chrome is open (Kortex MCP workflows docs, 2026). See our [daily briefing guide](/blog/gemini-notebook-daily-briefing) and [multi-step workflows guide](/blog/gemini-notebook-ai-workflows) for both.
## Start with one output
When we listed the production server's tools on October 1, 2026, it returned 150 of them, and 18 sit in the Studio and podcasts group alone. You don't need to learn them, though. Instead, ask for an output and let the assistant check, fetch and export.
Pro costs $9.99 a month, or $4.92 a month billed yearly (Kortex pricing, 2026), and Kortex also has a free plan. Pick the output you make most often and make it from chat next time.
Studio outputs from your chat, no tab shuffle
Start Audio Overviews, decks and quizzes from Claude or ChatGPT, then check, fetch and export them in the same chat. Kortex MCP is a private beta for invited Kortex Pro members.
## Frequently asked questions
### What is the Studio panel in NotebookLM?
Studio is the Gemini Notebook (formerly NotebookLM) panel where outputs based on your sources are made. Google's help lists notes, Audio Overviews, Video Overviews, mind maps, reports, data tables, flashcards, quizzes, slide decks and infographics. Kortex MCP, a private beta for invited Kortex Pro members, starts them from chat.
### Can Claude create podcasts like NotebookLM?
Yes, through Kortex MCP. Claude calls generate_audio_overview on your notebook with one of 4 formats (deep dive, brief, critique or debate) and a length, and Gemini Notebook (formerly NotebookLM) records it in your own Google account. Kortex MCP is a Kortex Pro feature in private beta, so you need an invite from the waitlist first.
### Can ChatGPT generate Gemini Notebook slides?
Yes. With developer mode on and Kortex added as an MCP app, ChatGPT can call generate_artifact with type slides, as a detailed deck or presenter slides, in short or default length. It can then revise chosen slides with revise_slide_deck and export the deck. Kortex MCP is invite-only while it's in beta.
### Does generating from chat use my Gemini Notebook limits?
Yes. Gemini Notebook makes every output in your own Google account, so it counts against Google's usage limits. Since September 2, 2026, those refresh every 5 hours until a weekly cap. Generating uses no Kortex podcast minutes; adding a finished episode to a Kortex podcast does.
### Can I export a Gemini Notebook quiz?
Not from Gemini Notebook itself. Google's help lists a CSV download for flashcards and none for quizzes. Through Kortex MCP, get_artifact_content returns every quiz question with its correct option, as JSON or one Markdown document, so your assistant can quiz you, grade answers or rewrite the set in another format.
### How do I get Audio Overviews into Spotify or Apple Podcasts?
Ask your assistant to add the finished Audio Overview to a Kortex podcast with add_podcast_episode, then get the RSS link with get_podcast_feed_url and add it by URL in your app. Open the Kortex dashboard once so the audio file is prepared. Podcasts are a Pro feature with 1,000 minutes included.
### Do I need to keep Gemini Notebook open?
No notebook tab is needed, but Chrome must be running with the Kortex extension signed in, because every call is relayed to it. The first call after Chrome has been idle can take up to 30 seconds while the extension wakes up. Later calls are fast.
---
*Editorial note: fact-checked against the official help pages and docs linked above on October 2, 2026. Spotted something out of date? [Contact us](/support) and we will fix it, or read more [about Kortex](/about).*