Self-hosted, no account required
No sign-up, no email, no workspace invite. The only identity involved is your computer's own user.
Air-gapped, open source NotebookLM alternative
A private research notebook that runs entirely on your own machine. Your documents, your model keys, no cloud, no account.
Trusted by experts at
How it works
Runs entirely on your machine
- 1
Your index lives on your disk
A file you own, not a row in someone else's database. With a local model, pull the network cable and it keeps answering questions about the sources you already added. That matters most for a case file or a client's ledger.
- 2
Your model, or your own key
Use SurfSense with a local model, or paste a key for a provider you already pay for. The key stays on the machine that uses it.
- 3
Sources become deliverables
Turn what you have indexed into a deck, a report, a briefing or a study guide, generated and stored in the same local database as everything else.
Why it is different
Three things NotebookLM cannot do
Bring your own model
NotebookLM gives you Gemini only. Here the model is a setting: a local Qwen or Gemma, or OpenAI, Anthropic and anything OpenAI-compatible.
Install and start: no Docker, no terminal
A normal installer for Windows, macOS and Linux, with llama.cpp built in. Pick a model on first launch and it runs on your computer.
How it compares
Getting started, compared
SurfSense
NotebookLM
- Runs offline
- Yes, fullyNo, cloud only
- Account required
- NoneGoogle account
- Setup
- Desktop installerWeb sign-in
- Choice of model
- Any, local or hostedGemini only
- Model bundled
- Picked at first launchNot applicable
- Source code
- Open, auditableClosed
What you get
Three things worth knowing before you install
Artifacts: decks, reports, briefings, podcasts
Every source can become more than an answer.
Twelve formats: Slides, reports, podcasts, quizzes and more from one set of sources.
Made offline: Generated and stored on your machine, with no per-minute fee.
Open source, extended by plugins
Each plugin gives the agent live data from one platform. Eight ship today.
More places to search: Google Search, YouTube, TikTok, Google Maps, Amazon, Walmart, Indeed and the open web.
Open by pull request: Our plugins are code in the public repo; anyone can list an MCP server in the registry.
Private by construction
Privacy as a structure, not a policy that can change.
Nothing to hand over: Data that never left your machine cannot be subpoenaed, breached or repriced.
No account, ever: Your index lives in your user folder. Delete the folder and it is gone.
For confidential work
Private AI for business, with no vendor holding a copy
Lawyers, accountants, consultants and engineers all have files they are not allowed to upload. SurfSense turns them into a deck, a report or a briefing anyway. The deliverable is built on the same machine the source sits on, so no vendor ever holds a copy of the contract, the ledger or the inspection report.
Learn moreArtifacts
Twelve things one set of sources can become
Listen
Podcast
Your sources talked through as an audio conversation.
Present
Slides
An editable presentation deck.
Present
Infographic
Key figures laid out as one visual.
Present
Image
Turn an idea into a picture.
Study
Mind map
Concepts laid out as a linked graph.
Study
Flashcards
A front-and-back deck for review.
Study
Quiz
Test yourself, with every answer explained.
Read and export
Markdown
A short brief of everything you have indexed.
Word
An editable write-up you can revise afterward.
PDF
A print-ready export of the same content.
Web page
A standalone HTML page.
Spreadsheet
Tables pulled out of your sources.
FAQ
Questions people ask
Can I self-host an AI?
Yes. A self-hosted AI runs the model and stores its data on hardware you control rather than a provider's servers. SurfSense is a desktop application that does this without a server at all: it installs like any other app, keeps its index in your user directory, and can run a local model so that no request ever leaves the machine.
Is there an AI I can use without internet?
Yes. SurfSense installs with llama.cpp built in, and its one-time setup screen has you pick a small language model that runs on the CPU. Once that one download finishes, you can add documents and ask questions with no network connection. An internet connection is only needed for that initial model download, if you choose a hosted model such as GPT or Claude, or when you download an update.
Can I run NotebookLM locally?
No. NotebookLM is a Google cloud service: it requires a Google account, uploads your sources to Google's servers, and cannot be installed on your own machine. Running the same workflow locally means using an alternative built for it, such as SurfSense, which keeps sources and answers on your computer.
Is there an open-source alternative to NotebookLM?
Yes. SurfSense is an open-source NotebookLM alternative: the source is public and auditable, it runs air-gapped on Windows, macOS and Linux, it works with any model you choose rather than Gemini alone, and it turns your sources into summaries, study guides and podcasts entirely offline.
Is there a free version of NotebookLM?
NotebookLM has a free tier with usage limits and requires a Google account. SurfSense is free and open source with no account, no quota and no trial: you download it, install it, pick a local model in the one-time setup screen, and chat at no cost. You only pay a provider if you choose to connect a hosted model of your own.