Air-gapped, privacy-focused, open-source NotebookLM alternative. AI agent that researches, transforms, and edits your documents. Join our Discord: https://discord.gg/ejRNvftDp9

SurfSense – Offline AI‑powered notebook

What it is – SurfSense is an open‑source desktop application that brings a Notebook‑style AI agent to your own computer. It lets you research, transform, and edit a wide range of document formats without sending any data to the cloud. You keep the index on disk, choose any OpenAI‑compatible or locally‑run model, and work completely air‑gapped.

Core capabilities

Capability	How SurfSense does it
Document research	The agent reads the content of PDFs, Word, Excel, PowerPoint, HTML, images (via OCR), audio recordings (via local transcription), etc., and returns cited answers that reference the original source.
Creation	Generate new documents (e.g., a Markdown report, a PowerPoint deck, or an audio podcast) directly from prompts or from a template you provide.
Editing	Ask the agent to modify a file – add text, change a table, insert comments, or apply tracked changes – and it writes the result to a new copy, leaving the original untouched.
Transformation	Convert between formats (PDF ↔ Word/Excel/PowerPoint, split/merge PDFs, extract pages, turn spreadsheets into pandas analyses, etc.).
Plugins	Built‑in plugins let the agent pull public web data (Google Search, YouTube, TikTok, Maps, Amazon, Walmart, Indeed) or third‑party services via a plugin registry.

Supported file types

PDF – research (including scanned pages via OCR), fill forms, add watermarks, split/merge.
Word (.docx) – research, generate from templates, edit with tracked changes, export to PDF.
Excel (.xlsx) – research formulas, generate sheets with charts, edit cells, analyse with pandas.
PowerPoint (.pptx) – research, generate slides, edit content, export to PDF.
CSV, Markdown, plain text, HTML – research and generate.
Images (PNG, JPEG, TIFF, BMP, WebP) – OCR for text extraction; can be used as input for image‑generation models.
Audio (WAV, MP3, FLAC, Ogg) – local transcription for research.
Podcasts, mind maps, flashcards – generation capabilities.

How it works

Install – Download the platform‑specific installer (Windows exe, macOS dmg, Linux AppImage or .deb).
Onboarding – Choose a language model: any OpenAI‑compatible API key or a locally‑run model (e.g., Ollama, Llama.cpp).
Add sources – Drag files or folders into the Sources panel; the app builds a local index.
Ask – Type a natural‑language query. The agent searches the index, optionally calls plugins for live web data, and returns a response with citations.
Edit/Transform – Use follow‑up commands to modify the original files; the app writes the result to a new file, preserving the source.

All processing (including OCR and transcription) can be performed on the machine; remote models are only used for the LLM inference step if you opt for them.

Platforms & licensing

Windows x64, macOS Apple Silicon, Linux x64 (AppImage & .deb).
License: Apache‑2.0 (permissive open source).

Comparison snapshot

Feature	Google NotebookLM (Gemini)	SurfSense
Runs offline	No	Yes
Documents leave machine	Yes	No (unless you use a remote model)
Can edit files	No	Yes
Open source	No	Yes
Model choice	Gemini only	Any OpenAI‑compatible API or local model

Getting started & community

Download: https://www.surfsense.com/downloads
Docs & dev loop: docs/README.md and the surfsense_local/ source tree.
Support: Discord server, GitHub Issues, and Discussions.
Contributing: PRs target the dev branch; see CONTRIBUTING.md.

SurfSense lets you keep sensitive documents truly private while still benefiting from AI‑driven research and editing – all on your own hardware.

MODSetter/SurfSense