Jack Roberts

Hermes Agent has a NEW SuperPower (NotebookLM)

⏱ 16 min video · 3 min read18 May 2026Worth watching

TL;DR

Jack Roberts demonstrates how to integrate Google's NotebookLM (free AI research platform) directly into the Hermes AI agent via Telegram, creating a 24/7 mobile research and action system. The integration lets users query curated knowledge bases, generate infographics and podcasts, and trigger automated workflows like daily email digests — all from a phone.

Key points

NotebookLM can be connected to Hermes agent using an unofficial Google skill, allowing programmatic control of NotebookLM from Claude Code and Hermes via a browser cookie authentication method.

Once integrated, users can query any NotebookLM notebook (up to 300 sources each) from Telegram via voice note or text message, returning summaries, infographics, and audio overviews at $0 cost.

The integration enables a read-to-action pipeline: NotebookLM provides the knowledge, Hermes acts on it — drafting emails, scheduling reminders, or triggering workflows without switching apps.

n8n MCP integration is demonstrated as a way to build visual, auditable automation workflows (e.g. a daily 8 a.m. email containing a YouTube strategy insight pulled from NotebookLM), giving granular control over non-deterministic AI outputs.

The setup requires downloading a NotebookLM skill file (free resource linked in description), uploading it to Claude Code, authenticating via browser, then adding the skill prompt to Hermes — described as a few-step process taking minutes.

Actionable insights

→

Download the free NotebookLM skill file from the video description, upload it to Claude Code, and authenticate by opening NotebookLM in a browser to capture the session cookie — this bridges NotebookLM to your local AI environment.

→

Use the Hermes Pantheon dashboard to create a dedicated NotebookLM skill persona with a system prompt, making it easy to visually manage and edit the skill later without touching raw config files.

→

For email-sending workflows, build them in n8n rather than asking Hermes to send emails directly — the visual execution log lets you audit outputs from non-deterministic AI before they reach recipients.

→

Build topic-specific notebooks with 250-300 sources and query them via Telegram voice note while on the go — NotebookLM handles all RAG retrieval at no cost, keeping your per-query expenses near zero.

Notable quotes

“You can have a vector database of 300 separate YouTube videos on anything you want and we can query it at $0 because we send the question and NotebookLM does all the heavy lifting and the retrieval and the RAG and it just spits back the answer.”

“NotebookLM answers, Hermes acts — you can draft scripts, save the notes, schedule reminders, ship the email, all from the same chat.”

“I could say every morning 8 a.m. go ahead and create me a NotebookLM on something that you think is going to be valuable based on our conversations in the last 24 hours — that is literally what we can do now.”

Worth watching?

✅

Worth watching the full video?

Watch if you actively use Hermes agent or Claude Code and want a step-by-step walkthrough of the setup — the key concepts and steps are all captured here, but the screen-share visuals of the Pantheon dashboard and n8n workflow builder may help if you get stuck on configuration.

Topics

AI & Tech NotebookLM

Explore more summaries on these topics →

Saved you some time? The creator still deserves a like.

Watch on YouTube →

More like this