open-notebook

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

• set up a research notebook
• chat with my documents
• analyze these papers
• NotebookLM alternative
• ingest research materials

What it does

open-notebook is a Claude Code skill from K-Dense AI’s scientific-agent-skills repo. It turns Claude into a setup and operation guide for a self-hosted, open-source NotebookLM alternative — allowing you to organize research materials into notebooks, ingest diverse content (PDFs, YouTube videos, audio files, web pages, Office documents), and run AI-powered analysis using any of 16+ providers including Anthropic, OpenAI, Google, Ollama, Groq, and Mistral.

A session produces a configured notebook environment where you can chat with your documents, generate AI-powered summaries, create multi-speaker podcasts from research papers, and search across all materials with full-text and vector search — with complete data privacy through self-hosting.

When to use it

Reach for it when:

• You want to run a local, private NotebookLM-style environment over your research document collection without sending data to Google
• You’re managing a large literature collection and want semantic search plus AI-generated summaries that you control
• You want to generate audio summaries or multi-speaker podcast episodes from your research papers for audio consumption

When not to reach for it:

• You just need a quick one-off literature search — use paper-lookup or literature-review
• You’re managing references in Zotero — use pyzotero for programmatic library access

Install

Copy the SKILL.md from K-Dense AI’s open-notebook folder into .claude/skills/open-notebook/ in your project. The skill guides you through setting up the self-hosted open-notebook service and configuring your chosen AI provider credentials.

Trigger phrases: “set up a research notebook”, “chat with my documents”, “analyze these papers”, “NotebookLM alternative”.

What a session looks like

A typical session has three phases:

1. Environment setup. Claude walks through installing and configuring the self-hosted notebook service, selecting an AI provider (or multiple), and setting up the vector search index.
2. Content ingestion. You point Claude at your research materials — a folder of PDFs, a list of URLs, YouTube links, or uploaded documents. The ingestion pipeline handles format detection, text extraction, chunking, and embedding.
3. Active use. With materials ingested, you can ask questions across the entire corpus, generate structured summaries, request podcast scripts from selected papers, or run custom content transformations — all against your private data.

Receipts

Where it works well:

• Literature collections where you want semantic search plus AI synthesis — finding the three papers most relevant to a specific mechanism across 200 PDFs takes seconds rather than manual scanning
• Privacy-sensitive research data that can’t leave your institution — self-hosting means your data doesn’t transit third-party servers

Where it backfires:

• Initial setup takes longer than a SaaS tool — if you only need to process a handful of documents once, the setup cost doesn’t pay off
• Podcast generation and audio features require additional TTS provider credentials on top of the base LLM setup

Pattern that works: set up the notebook once for a research project and keep it running; the value compounds as you add more papers and the corpus becomes searchable rather than staying as isolated PDFs.

Source and attribution

Originally authored by K-Dense Inc.. The canonical SKILL.md lives in the open-notebook folder of their public scientific-agent-skills repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner’s perspective. For the formal spec and any updates, defer to the source repo.