Mervadon Feature Overview: A More Reliable Way to Bring Web Research into NotebookLM

When you use NotebookLM for research, asking questions is often not the most time-consuming part. Preparing the sources comes first, and that is where much of the work happens.

You may find yourself moving between a dozen web pages and copying links into NotebookLM one by one. Sometimes you only need two paragraphs from a long article, but end up importing the entire page. As your source collection grows, new questions appear: Which page has already been imported? Which selection is still pending? Which Google account is the current operation using?

Mervadon is designed to handle this often-overlooked stage before the work reaches NotebookLM.

It is a Chrome extension that connects web research with NotebookLM. You can capture full pages or precise text selections, save and organize them locally, and then import the material you need into a specific Notebook. Mervadon can also synchronize sources and artifacts, while letting you reuse frequently used generation prompts and parameters.

Between versions 0.1.37 and 0.1.47, Mervadon did more than add a few new entry points. It developed a more complete workflow for capturing, connecting, importing, and verifying research material. Here is what the current feature set looks like in practice.

1. Connect to NotebookLM in the background without keeping a tab open

The most significant addition in this release range is the NotebookLM background Direct connection.

After you sign in to NotebookLM and grant the required site access, Mervadon can use the browser's existing signed-in session to read your Notebook list, synchronize Sources and Artifacts, and carry out user-initiated source imports, deletions, and artifact generation in the background.

For routine research work, this means you generally no longer need to keep a NotebookLM tab open just to maintain the connection. You also do not need to configure a separate AI provider, enter an API key, or export cookies manually.

The workflow becomes more straightforward:

1. Sign in to NotebookLM in your browser.
2. Grant access and connect through Mervadon.
3. Select the target Notebook.
4. Continue reading and capturing material on the web.
5. Let the background process handle list synchronization and import operations.

After you grant NotebookLM access for the first time, the background process continues by validating the existing session and loading the Notebook list. If Chrome's permission prompt temporarily closes the Popup, you do not need to click Connect again.

There is an important boundary here. The background connection relies on an existing, verifiable NotebookLM session. It does not bypass authentication, permissions, or NotebookLM service limits. If the session expires, you still need to sign in again. If Mervadon cannot verify the account identity, it will not treat an untrusted account state as a successful connection.

2. Import the exact selection you need, not the entire page

A common research scenario is finding one useful definition, data point, or argument inside a long article.

Mervadon's precise selection import is built for this situation.

First, choose the target Notebook in the Popup. Then select text on the page and start the capture through the Capture selection toolbar action, the context menu, or the Popup. Mervadon saves the selected text locally before importing it into the target Notebook as a text source.

The important word is "precise":

• The import contains only the current selection, without mixing in the full page text.
• Mervadon does not substitute a full-page URL source for a selection import.
• Different selections from the same page are not treated as duplicates merely because they share a title or URL.
• Repeated submissions of identical content reuse an existing operation or receipt when possible, reducing duplicate sources.
• If a selection exceeds the single-import text limit, the full text remains saved locally and Mervadon asks you to shorten the selection instead of silently truncating it.

This is particularly useful for research-heavy work. Rather than sending a several-thousand-word article into NotebookLM, you can import only the material that directly supports the question you are investigating. That reduces source noise and makes later verification easier.

If no target Notebook is available, Mervadon keeps the content locally and asks you to select one. It does not present a local save as a completed remote import. The interface reports a successful import only after it receives a receipt that matches the target Notebook and source content, and confirms that the remote source is ready.

3. Capture permissions remain an explicit user choice

A browser extension that reads page content needs clear permission behavior.

Mervadon now uses the same capture-permission flow across the Popup, Workspace, and Settings page. On first use, you can explicitly choose between:

• Granting access only to the current website.
• Granting access to all standard HTTP and HTTPS websites.

If you have not made a choice, Mervadon does not interpret an unset state as consent to broad access. Web capture permissions and NotebookLM connection permissions are handled separately. Connecting to NotebookLM does not automatically request access to every regular website.

The post-permission workflow has also been improved. After access is granted, the original capture request can continue, and previously granted permission is reused so that the same operation does not trigger repeated prompts. If the Popup closes and is reopened, a saved request is not automatically treated as a new submission.

As of version 0.1.46, the Workspace's default optional-permission request list no longer includes identity or identity.email, reducing the need to request additional browser identity permissions by default.

If permission is denied, the system does not report a successful capture or import. This may look like a small messaging detail, but it follows a basic rule: the interface should reflect the operation's actual state.

4. Save frequently used generation settings as templates

Mervadon also addresses another repetitive task: re-entering prompts and parameters every time you generate an artifact.

Since version 0.1.37, users have been able to manage templates by artifact type from the Settings page. A template can store its name, description, prompt, and the generation parameters supported by that artifact type. The management interface covers audio, video, mind maps, reports, data tables, flashcards, quizzes, slide decks, and infographics. The parameters you can edit depend on the parameter set available for each type.

For example, you can save a fixed structure for a team report, set a consistent language for study materials, or preserve a recurring prompt for repeated research tasks. When generating an artifact in the Workspace, you can select the relevant template directly. If no template is configured, Mervadon uses the default parameters, so you do not need to create a full configuration before generating something.

Templates are stored per NotebookLM account and can be exported as JSON for backup and inspection.

Template management does not mean that Mervadon introduced all of the listed NotebookLM artifact types, nor does it mean template-file import is currently available. The feature is specifically for reusing parameters within the existing generation workflow.

5. Generate reports in an Interactive Learning Overview format

If you use NotebookLM to organize course material, academic papers, or topic-based research, Mervadon also provides the interactive_learning_overview report format.

Using a dedicated learning-overview prompt, it organizes key concepts, summaries, and suggested learning activities, turning a set of sources into material better suited to guided learning and review.

This is one Report format. It is not a promise to create several independent Studio artifacts in a single operation. The actual output still depends on NotebookLM, the selected sources, and the generated result.

6. Clearer account identity in multi-account workflows

When you use both personal and work accounts, the real risk is not an extra click. It is importing material into the wrong Notebook.

Mervadon has made several targeted improvements to its multi-account workflow. It can identify an email address from Google's official account controls while also checking that the account index, email address, and session identity agree. If the information conflicts or the session is invalid, Mervadon refuses to treat it as a connected state.

The existing account selector is now integrated into the account card in the Popup. You can switch accounts by clicking the card, using the dropdown arrow, or navigating with the keyboard. If a switch fails, the interface restores the original account instead of remaining in a state that looks updated but has not actually taken effect.

Notebooks, Sources, Artifacts, and related operations remain isolated by account. This prevents cached data or import receipts from another account from being applied to the current target.

7. See usable content first while remote data refreshes

A common problem with connected tools is opening the interface only to face a blank screen or a long loading state. Between versions 0.1.40 and 0.1.44, Mervadon reworked the Popup's loading and refresh behavior.

If the account, Notebook list, and selected target are already cached locally, the Popup displays them first and refreshes remote state in the background. If the network is slow or the refresh fails, usable cached data remains available. When asynchronous requests return, Mervadon checks the current view so that a result from an older account or request cannot overwrite a newer selection.

The first connection still displays a clear loading message and spinner, avoiding the misleading appearance of a disconnected state or an empty list while data is still being read. When fresh cache data is already available, Mervadon does not repeatedly show the longer first-connection message.

For the current account, the automatic flow may reuse cached data when the last successful synchronization happened less than five minutes ago and access permission is still valid. A manual refresh always performs a real refresh and displays progress. If it fails, the controls become available again, existing data is preserved, and the interface reports the reason.

These changes do not make the network faster. They make it clearer whether Mervadon is loading data, reusing cache, or handling an actual error.

8. After an import error, verify the state before submitting again

The hardest import failures are not explicit successes or failures. They occur when a request may have been submitted but the subsequent status check fails. Retrying immediately in that situation can create duplicate sources in NotebookLM.

Version 0.1.47 focuses on recovery from this uncertain state.

For network errors, timeouts, and temporary service failures, Mervadon performs a limited retry when reading the Source list, with no more than two reads during a single verification attempt. It does not replay the source-write request. Authentication, permission, and rate-limit errors are not automatically retried as if they were ordinary network interruptions.

Error messages also preserve the underlying cause when possible, including available HTTP status codes or RPC identifiers. This makes it easier to distinguish an expired login from a service failure or protocol problem.

The receipt sequence is especially important. Once Mervadon receives a remote source ID, it first saves a receipt tied to the original content and then checks whether the source is ready. If that later read fails, the interface reports the source as submitted and awaiting confirmation rather than declaring it complete too early.

If the background process is interrupted, a local receipt fails to save, or post-submission verification encounters an error, the next attempt first checks the original operation instead of immediately importing the source again. A previous failure record that confirms no submission occurred can be checked again during the user's next explicit attempt, rather than permanently blocking the workflow.

For the user, the principle is simple: when the state is uncertain, verify what happened before deciding whether to retry.

A more practical Mervadon workflow

Taken together, these features support a complete workflow like this:

1. Sign in to NotebookLM in your browser and grant Mervadon access to NotebookLM.
2. Confirm the current account in the Popup and choose the target Notebook.
3. Grant capture access to the current website or to standard websites, depending on your needs.
4. Capture a complete page or select only the relevant passage while reading.
5. Review the locally saved content before starting the remote import.
6. Decide what to do next based on the actual state, such as awaiting confirmation, submitted, or ready.
7. When generating an artifact, select a saved prompt template or use the default parameters.
8. Return to the Workspace to view Sources and Artifacts, synchronize data, or inspect the output.

The point is not simply to reduce the number of clicks. It is to make every step inspectable: what was captured, where it will be imported, which account is active, whether the remote service received it, and whether it is ready.

Who is Mervadon for?

If you only add one or two links to NotebookLM occasionally, NotebookLM on its own may already be enough.

Mervadon is more useful for people who collect research material continuously: research writers, content creators, analysts, consultants, students, educators, and anyone who needs to preserve useful evidence from large numbers of web pages or AI conversations.

It does not replace NotebookLM, and it does not decide whether a source is trustworthy. It fills in a more fundamental part of the research workflow: saving pages and selections accurately, delivering them to a clearly identified destination, and leaving enough operational context to recover when something goes wrong.

From version 0.1.37 through 0.1.47, Mervadon brought background connectivity, precise selection imports, permission onboarding, template management, multi-account handling, cache-aware refreshes, and import recovery into a more complete workflow.

For frequent NotebookLM users, these changes address the details that appear in everyday work: less manual transfer, less uncertainty about status, and fewer accidental or duplicate imports.

Product: https://mervadon.com/
Chrome Web Store: https://chromewebstore.google.com/detail/mervadon-notebooklm-works/igjlegcbacndfchdlganecomklaocjin

Note: This article is based on local Mervadon builds and feature records from versions 0.1.37 through 0.1.47. The version numbers do not mean that every individual version was released publicly through the Chrome Web Store. Actual feature availability depends on the installed version, the user's NotebookLM session, and NotebookLM service capabilities.