Claude Code + NotebookLM Automatizam a minha vida! (Nova Skill)

Main Idea / Use Case

The video explains how to automate research and content preparation by combining:

1. Claude Code (described as an “AI Jarvis” / general-purpose agent builder)
2. NotebookLM (Google’s “university of knowledge” tool for summarizing and extracting information from content)

This workflow is aimed at people who do daily research/study or create content (e.g., YouTube/Instagram). The goal is to save time on:

• finding sources,
• summarizing them,
• generating presentation-ready materials such as slides and infographics.

Key Technological Concepts & Product Features

1) Claude Code (agents / skills)

Claude Code is presented as a general-purpose AI that can connect to other tools when given the right context, enabling “systems” and “applications.”

Key focus: skills inside Claude Code.

• Skills are described as autonomous agents that automate business tasks.
• Skills can improve over time based on additional context you provide.

Claude Code can orchestrate multi-step pipelines, such as:

• YouTube search → select a video (often the one with the most views)
• send the chosen content to NotebookLM
• generate a structured 5-step infographic / slide-style output

2) NotebookLM (end-to-end content understanding)

NotebookLM is described as an end-to-end Google product that supports:

• content research
• scraping/transcription (turning audio/video transcripts into analyzable text)
• rapid summarization (e.g., a 40–120 minute podcast/video condensed into ~1.5–2 minutes, or even ~30 seconds)

It also functions as a context library / AI knowledge base:

• stores and retrieves knowledge from the provided content
• allows Claude Code to run more automated pipelines

The video also claims NotebookLM can generate new content formats derived from video input—e.g., turning watched videos into podcast-like outputs via AI.

Automation Workflow Demonstrated (Claude Code + NotebookLM + YouTube)

An example workflow:

1. The agent searches YouTube for a topic (e.g., “Claude Code Skills 2.0” or “Claude Code sub-agents”).
2. It chooses the most relevant video—often based on highest view count.
3. The video content is opened in NotebookLM.
4. An infographic is generated in a specific style, including requirements such as:
   • Brazilian Portuguese (PT-BR)
   • 16x9 format (computer full-screen)
   • white background with blue elements
   • handwritten typography (described as “written by hand,” aligned with subtitle style)

Tools / Environment Integrations Mentioned

Antigraft (visual interface)

Antigraft is described as a visual tool/browser-like UI for navigating and viewing Claude Code outputs.

The video compares it to:

• a browser UI, and
• a database that renders code/files visually (e.g., Python becomes visual “cards”).

Additional claimed features:

• Antigraft reportedly includes Gemini 3.0 / 3.1 for extra capabilities like design generation without external tools.

The presenter uses Claude Code via terminal, then views outputs inside Antigraft.

Tanglin / API access claim (installation blocker)

The video claims NotebookLM does not provide an API key in the way some automation tools (e.g., n8n) expect.

It credits Tanglin with enabling the integration via an approach that works around this limitation.

The presenter contrasts:

• “DIY n8n workflows” (scrape transcripts → manage infrastructure → store large knowledge bases), which the presenter calls heavy
• versus using Claude Code + NotebookLM, where NotebookLM provides the “knowledge-base behavior” without building everything from scratch.

Step-by-Step Setup (Installation / Tutorial Elements)

The video includes an installation flow for enabling skills in Claude Code.

Install NotebookLM Skill

1. Copy/paste the first installation guide command into Claude Code terminal.
2. Copy/paste the required CLI.
3. The CLI opens a Google browser popup to connect your NotebookLM account.
4. Add the Python API so the agent has context on how NotebookLM API requests work.
5. Download/import/copy the agent skills so the setup installs the NotebookLM integration.

The presenter claims this takes ~2–3 minutes.

Style/context examples included:

• Provide an infographic template reference so future images match that template.
• Provide brand identity context (e.g., white background, blue elements, typography) so output is consistent.

YouTube Search Skill Setup

The YouTube Search skill is described as already available as an MD/Docs-based setup:

• Open Google Docs
• copy the text
• create a new text file in Antigraft
• save with Ctrl/Cmd+S
• install/use the skill (credited to Chase AI as the creator)

Review / Analysis Points Emphasized

Competitive advantage vs DIY automation

• Building a similar system in n8n is described as difficult due to NotebookLM API limitations and heavy infrastructure needs (transcript scraping, large storage, retrieving relevant video segments).
• NotebookLM is described as a “great library/university” that simplifies context handling.

Cost/credits advantage

• The video claims the two skills (YouTube Search + NotebookLM) can be used for free with Claude Code, “without credits,” including avoiding payment for:
  • infographic generation
  • YouTube searching (as stated in subtitles)

Who Is Speaking / Sources (as mentioned)

• Main speaker / presenter: the channel host (referred to as “I” throughout), who invites viewers to subscribe and discusses their daily workflow and business.
• Credited contributors:
  • Tanglin (enabling the NotebookLM integration workaround/skills)
  • Chase AI (created/published the YouTube Search skill setup)
  • Rafael Quintanilha (source of referenced videos on skills/MCPs/sub-agents)
  • Full Cycle Practice (referenced for sub-agents content)