SurfSense

Air-gapped local index The search index stays on your own disk and documents leave the machine only when you pick a remote model. This makes SurfSense usable for files that policy or prudence forbids uploading to any cloud notebook.[1]

Local or remote model choice You pick the model: any OpenAI-compatible API or a fully local one. Nothing about the app locks you to a single vendor's model, unlike closed notebooks that only run their own.[1][2]

Cited research across formats Ask questions with cited answers across PDF including scanned pages, Word, Excel with formulas, PowerPoint, CSV, images via OCR, Markdown, text, and HTML. The agent also analyses spreadsheets with pandas and charts the results.[1]

Editing that never touches originals Edits, conversions, and PDF tools always produce a new file while the original stays unchanged. Word body text becomes tracked changes and comments, Excel cell values and formulas are edited in place in the copy, and PDFs gain filled forms, watermarks, page numbers, headers, and footers.[1]

PDF and Office transformations Merge, split, extract, rotate, and reorder PDF pages, convert Word and PowerPoint to PDF with Office support, generate Word and Excel files with formulas and native charts, and create podcasts with on-machine voiced transcripts.[1]

Desktop installers, no account Installers ship for Windows x64, macOS Apple Silicon, and Linux x64 as AppImage and deb, with onboarding setup inside the app and no account to create. Mind maps, flashcards, and quizzes are generated in the app.[1]

Legacy connectors and channels The documented 0.0.40 stack connects Notion, Slack, Google, Jira, and more through built-in scraper APIs, runs local models via Ollama and LM Studio, and reaches chat through Telegram, WhatsApp, Slack, and Discord. These docs describe version 0.0.40 only and the 2.0 docs are still being written.[1]

Free app plus plugin licence The app and every update are free forever with no usage cap, and a 30-day plugin licence arrives by email with no card. The paid licence unlocks nine platform scraper plugins plus the web crawler at a flat price with no per-item metering, and the first plugin ships through the coming in-app plugin system.[1]

Anyone handling documents that must never leave the building, such as legal matter files, patient-adjacent records, or unpublished research, gets a NotebookLM-style workflow without the upload step that rules out closed alternatives.[1][2]

People who already run local models or hold their own API keys will like choosing either per setup, keeping costs at zero for the app itself and paying only for model usage they already control.[1][2]

Readers who need more than chat, such as filling PDF forms, stamping page numbers, editing Word with tracked changes, or charting spreadsheet data, get document work done where pure Q&A notebooks stop.[1]

The hosted cloud service is export-only and every stored file is deleted on 18 October 2026. Anyone with workspaces there must export the ZIP and import it into the desktop app before then; original uploads, generated artifacts, and live citation links do not carry over.[1]

Web scraping connectors cost extra: nine platform plugins plus the web crawler need the paid licence, and no plugin ships in version 2.0 yet. The first plugin will be the hosted scraper API installed from inside the app once the plugin system is ready.[1]

The repository mixes licences: most content is Apache-2.0, but everything under surfsense_backend/app/proprietary/ sits under Business Source License 1.1. Adopters with strict open-source policies should review that directory before building on the code.[1]

No mobile app, browser client, or multi-user hosted workspace is documented for version 2.x. Households or teams that need shared access from phones or a browser should look at a server-based notebook such as Open Notebook rather than waiting for clients that are not announced.[1][2]