Open Notebook
Self-hosted research notebook where you chat with your sources, make notes and generate multi-speaker podcasts.
Chat with PDFs, videos and web pages using the AI models you choose.
Open Notebook is a self-hosted alternative to Google's NotebookLM. It suits researchers and students who want to ask questions across their sources, with citations, while choosing which AI provider sees their material.
What it is
You add PDFs, videos, audio, web pages and Office documents to notebooks, then chat with them, search by full text and vector, and write AI-assisted notes. It can turn a notebook into a podcast with one to four speakers, and it exposes a full REST API.
Who it's for
Suits researchers, students and privacy-minded professionals who are comfortable running Docker and supplying their own AI keys or a local model server. It does not suit anyone who wants a hosted app with nothing to install.
Key features
Choice of 20+ AI providers
Connect OpenAI, Anthropic, Google, Mistral, Ollama, LM Studio and others, and assign separate language and embedding models.
Podcast generation
Build podcasts with 1 to 4 speakers and custom Episode Profiles, where NotebookLM offers two speakers.
Cited chat
Answers cite the sources, notes and insights they drew on.
Custom transformations
Define your own reusable actions to summarise or extract insights from a source.
Fine-grained context control
Choose exactly what each chat shares with the AI model.
REST API and MCP
A full REST API and an MCP integration let other tools drive your notebooks.
Up and running in 3 steps
Our honest verdict
What's great
- You pick the model per task, including fully local ones via Ollama, so sensitive research need not leave your machine.
- Podcasts are not capped at two voices and you control the script through episode profiles.
- The REST API gives you programmatic access that the README says NotebookLM does not offer.
- The interface is offered in 13 languages, including Turkish, German, Japanese, Polish, Catalan and Bengali.
Where it falls short
- You run it yourself: Docker with two services (SurrealDB and the app), and you must set an encryption key before first start.
- The compose file publishes the UI and API on all network interfaces with no password by default, so on a shared network you must bind to localhost or set a password.
- You pay your AI provider per use, and Anthropic and Groq have no embedding models, so you need a second provider for embeddings.
- Running local models for chat and podcasts needs your own capable hardware, which the README does not size for you.
Pricing & license
Open Notebook is MIT-licensed and free to run. The README says you pay only for AI usage from the provider you choose, or nothing extra if you use local models. The project shows no paid hosted plan on its site.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Does Open Notebook send my documents to the cloud?
Only to the AI providers you configure. With a local model server such as Ollama, the README says nothing leaves your machine.
Do I need to pay for anything?
The software is MIT. You pay whichever AI provider you connect for usage, or run local models on your own hardware.
How does it compare with NotebookLM?
The README claims more provider choice, 1 to 4 podcast speakers instead of 2, a REST API and self-hosting, versus Google's hosted-only service.
