Research Engine: Claude Code, NotebookLM & Obsidian Integration

Traditional research is often a fragmented manual process of managing tabs, reading articles, and copy-pasting notes. This guide outlines how to build a unified, automated research pipeline that queries YouTube, runs deep analysis using NotebookLM, and stores results cleanly inside Obsidian.

The Architecture & Roles

• Claude Code (Execution): The command-line interface that runs tools, queries sources, and manages files directly inside your local vault.
• Skill Creator (Extensibility): A Claude Code utility to build custom scripts and shortcuts using natural language instructions.
• NotebookLM (Analysis): Google's AI research tool that performs heavy document parsing and summarization on Google's cloud infrastructure, saving your Claude API tokens.
• Obsidian (Memory): Your local markdown database. Over time, Claude Code reads your links and vault files to align research outputs with your formatting preferences.

Step-by-Step Integration

• Step 1 (Install Skill Creator): Initialize the customization interface by running skill-creator within your vault folder in Claude Code.
• Step 2 (YouTube Skill): Build the query engine using: /skill-creator I want to create a skill that searches YouTube (ensure yt-dlp is installed).
• Step 3 (Setup NotebookLM Link): Install the open-source library notebooklm-py via your main terminal, and run notebooklm login to authorize Google Account access.
• Step 4 (NotebookLM Skill): Teach Claude Code to control the API: /skill-creator create a skill to control notebooklm-py.
• Step 5 (Orchestration): Chain the actions together into one pipeline: /skill-creator create a YouTube research pipeline.

Execution & Results

• Trigger the pipeline with a single query: /yt-pipeline Research emerging AI agent frameworks in 2026.
• Background Pipeline: The system automatically extracts top video transcripts, pushes them to NotebookLM, performs analysis, and returns mindmaps/infographics.
• Outcome: In under 6 minutes, a complete markdown report is saved to your local vault, ready for linking.