Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Power Local LLM
RAG & Document Chat

SurfSense Review: Open-Source Desktop NotebookLM Alternative

·8 min readBy Hans Kuepper · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

SurfSense (version 2.2.2, released 16 October 2026) is a free, open-source desktop app that, per its own site, lets an AI agent research, transform and edit your documents on your machine, with no account and the index kept on disk. You choose the model yourself, and documents leave the computer only if you pick a remote one. Every privacy claim comes from the project; PromptQuorum has not tested the app or its network traffic.

Key Takeaways

  • Maintenance: v2.2.0, v2.2.1 and v2.2.2 were published on 9, 14 and 16 October 2026, so no stale-project warning applies.
  • License: Apache-2.0, with the full source on GitHub.
  • Cost: the app is free; only the project's own web plugins and an Enterprise tier are paid.
  • Hosted version: the earlier cloud service was scheduled for data deletion on 18 October 2026, so the desktop app is the only route.
  • Gate: the download form asks for an email address before it hands over the installer.

📍 In One Sentence

SurfSense is an Apache-2.0 desktop app that searches, cites and edits your own files with a model you pick.

💬 In Plain Terms

You install it, point it at files or folders and a model, and ask questions; the project says the index and your documents stay on your disk.

📌Note: This review is based on the SurfSense website, privacy policy, sunset page and GitHub release page, checked on 21 October 2026. PromptQuorum has not installed, tested or benchmarked the app.

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

SurfSense is an open-source desktop app that the project describes as a privacy-focused, NotebookLM-style AI agent for your own documents. It grew out of a hosted web service that the project has shut down. This review reflects its website, privacy policy and GitHub release page as checked on 21 October 2026.

SurfSense Review: Open-Source Desktop NotebookLM Alternative

What Is SurfSense?

SurfSense is a desktop app in which you add files or folders as sources and ask an AI agent to find, cite, extract, transform or edit what is in them. The project calls it a privacy-focused, NotebookLM-style agent and positions it against Google's hosted NotebookLM.

It replaces SurfSense's earlier hosted web app: the project moved to a local desktop app and, according to its sunset page, ended the cloud service. The repository is MODSetter/SurfSense on GitHub.

Get It

SurfSense is installed from the downloads page on its website; the page lists no mobile, web or Intel Mac build.

Platform
Where to get it
Windows (x64)Installer SurfSense-Setup.exe from the downloads page
macOS (Apple Silicon)Disk image SurfSense-arm64.dmg (1.26 GB per the release page)
Linux (x64)SurfSense.AppImage or SurfSense.deb from the downloads page
Websitesurfsense.com
Source codegithub.com/MODSetter/SurfSense
Privacy policysurfsense.com/privacy (updated 18 September 2026)

Version reviewed: v2.2.2, published on 16 October 2026 per the GitHub release page, which is the app's latest upstream update at the time of checking. The downloads page lists no system requirements. Known issues in the release notes: the first start needs free disk space roughly the size of the database file, and rolling back to 2.2.1 means restoring a database backup by hand.

Pricing

The app itself costs nothing; a paid Individual plan adds the project's own web plugins.

Plan
What it includes
Price (USD)
FreeApp and all updates, no account or usage cap, plus a 30-day plugin license for an email address$0
IndividualPlugins for Google, YouTube, TikTok, Maps, Amazon, Walmart, Indeed, web crawler, priority support$60 first year, then $120/year
EnterpriseSelf-hosted deployment inside your network, central controls, audit logging, named supportCustom, invoiced annually

Prices are from the pricing page, read on 21 October 2026. Third-party plugins and MCP servers need no license, and the license does not lock any of the app's own features.

How to Get Started

The steps follow the order on the project's site; PromptQuorum has not run them.

  1. 1
    Download the installer
    Why it matters: Choose the Windows, macOS or Linux file on the [downloads page](https://www.surfsense.com/downloads).
  2. 2
    Complete the onboarding and add a model
    Why it matters: Connect any OpenAI-compatible API or a local model. Image features need an image model, and none is selected by default.
  3. 3
    Add sources
    Why it matters: Drop files or folders into the sources panel.
  4. 4
    Ask the agent
    Why it matters: Ask it to search, extract, transform or edit what is in your sources.
  5. 5
    Optionally add plugins
    Why it matters: Open Settings › Plugins to add web plugins or MCP servers.

Features Confirmed by the Sources

Every item below comes from the project's website, release notes or privacy policy.

  • Cited research. Answers draw on your sources and cite them; since 2.2.2, PDFs with a text layer and Word files are ready in seconds because they no longer wait on Docling.
  • Audio. Recordings can be added as sources and are transcribed with on-device speech recognition; the chat box has a microphone for voice input.
  • Plugins. Under Settings › Plugins, web and platform plugins (search, crawling, video and shopping sites) extend the agent; they need internet access and the license described under Pricing.
  • MCP. Third-party MCP servers (the site names DeepWiki and Notion) can be added from the plugin registry or by URL and stay off until enabled.
  • Tool permissions. Each tool can be set to Always allowed, Needs approval or Blocked.
  • Office files and podcasts. Office support through LibreOffice is optional and switched on in Settings; podcasts are saved as MP3.

Privacy and Network Use

The privacy policy says the desktop app has no account or sign-in, keeps documents, chats, search index and settings as local files on your computer, and never sends them to SurfSense. It adds that the app has no analytics, usage tracking or crash reporting.

The policy names two other outbound connections, update checks to GitHub and model downloads from Hugging Face, both off until you enable them; license checks run offline. If you configure a remote model provider, your prompts and the relevant document excerpts go to that provider under its own policy. The website is different: it uses PostHog analytics by default with no consent banner and does not honour Do Not Track or Global Privacy Control signals, and Stripe handles payments.

📌Note: The project presents the app as able to run offline and air-gapped, but PromptQuorum has not audited the code or the network traffic, so these are the project's claims. Its own web plugins always need internet access.

If You Used the Hosted Version

The project's sunset page said the earlier cloud service would delete all data on 18 October 2026 and that the desktop app replaces it. Anyone holding an export can bring it across as follows.

  • Export contents. The ZIP holds ready documents as Markdown, the folder structure and chat threads; it does not include original uploads, generated artifacts, tool calls, agent steps or live citation links.
  • Import. In the desktop app, open the gear icon at the bottom of the left rail, go to General, find "Import from SurfSense cloud", click Upload and pick the ZIP; a fresh install also shows an Upload button on its empty workspace screen.
  • After the import. Workspaces, folders and chat threads appear, and the app indexes the documents on your machine.
  • Deadline passed. PromptQuorum could not confirm whether exporting is still possible; check the sunset page.

Who Should Use It

  • People who want NotebookLM-style research without uploading files. Per the project, documents stay on disk unless you choose a remote model.
  • Users who already run a local or OpenAI-compatible model and want an agent front end for it.
  • Tinkerers who want to read or build the code from the GitHub repository.

What We Could Not Verify

  • Hands-on quality. PromptQuorum did not run the app, so answer quality, citation accuracy, speed and the agent's file edits are not assessed.
  • Offline and telemetry claims. The air-gapped and no-telemetry statements were not checked against the code or network traffic.
  • Hardware needs. The sources state no minimum RAM, disk or GPU.
  • Self-hosting. The repository has a docker folder, but the pages checked give no self-hosting steps, so the desktop app is the only documented free route.

Competitors and Alternatives

PrivateGPT

Platforms:
Windows, macOS, Linux (web UI, CLI)
Price / license:
Free / Apache 2.0
Key difference:
Offline-by-design RAG aimed at air-gapped setups, run from a terminal
Articles about PrivateGPT (8)

Also mentioned in:

+8 more not shown

HilbertRaum

Platforms:
Windows, macOS (Apple Silicon), Linux
Price / license:
Free / GPL-3.0-or-later
Key difference:
Portable no-install desktop workspace for offline chat and document Q&A
Articles about HilbertRaum (2)

Competitor details change often; confirm each app's current price, license and platforms on its own listing.

Compare in directory

Frequently Asked Questions

Does SurfSense need an account?

No. There is no sign-in; the only personal detail the policy names is the email address used for the download form and the trial license.

Can I run it fully offline?

Per the project, yes once a local model is set up. The web plugins cannot work offline because they read public pages. PromptQuorum has not tested this.

Does it work with ChatGPT or other providers?

The project names any OpenAI-compatible API, and release 2.2.2 adds sign-in with a ChatGPT Plus or Pro plan. No other provider is named in the sources checked.

How is it different from NotebookLM?

The project's own comparison table says NotebookLM supports only Gemini, while SurfSense accepts any OpenAI-compatible API or a local model and can run offline. PromptQuorum has not checked NotebookLM's current features.

Verdict

SurfSense is a free, Apache-2.0 desktop app that, per the project, keeps a document agent on your own machine with a model of your choice, and it is actively updated, with three releases in the week before this check. Against that, its privacy and offline claims are unverified here, it has no Intel Mac or mobile build, the plugins that read websites cost money and the hosted version it grew out of is gone. It suits people who want a private, NotebookLM-style workflow and are comfortable connecting their own model; readers who prefer a longer-established tool can compare AnythingLLM or PrivateGPT.

Sources

← Back to Power Local LLM

MCP