Product / UX / UI & API

SurfSense

Open-source, privacy-focused AI research assistant positioned as a self-hostable alternative to NotebookLM, Perplexity, and Glean. It connects an LLM to your personal or team knowledge base plus a wide range of external sources and lets you chat over them with hybrid RAG retrieval. The differentiator is unlimited sources and notebooks with no data caps, full freedom to bring your own LLM/embedding/reranker models (including local ones), and a deep connector catalog - all self-hostable so data never leaves your infrastructure.

Apache-2.0; ~15k GitHub stars by mid-2026. Self-host via Docker/Compose (PostgreSQL + pgvector). Connector count and per-feature depth are taken from the maintainer-authored README and not independently verified. Newer/smaller project than the leading open chat UIs.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public(github.com/MODSetter/SurfSense)
self-host: primary(Docker/Compose)
core-gated: ungated

Permissive OSI license (Apache-2.0), complete source public, self-host-first via Docker/Compose. No paid tier or gated features found on the repo or site (mild uncertainty - treated as fully open until a commercial tier appears).

https://github.com/MODSetter/SurfSense/blob/main/LICENSE recorded 2026-06-17
Apache License 2.0
https://github.com/MODSetter/SurfSense recorded 2026-06-17
~15k stars; open privacy-focused NotebookLM/Perplexity alternative; Docker self-host

Adoption

3 medium confidence

3.0

~15k GitHub stars by mid-2026; no download/usage telemetry available for a self-hosted web app, so stars_fallback (capped at level 3). Rapid star growth but unproven real-world deployment.

https://github.com/MODSetter/SurfSense recorded 2026-06-17
~14.9k stars, ~1.4k forks

Capability

4 low confidence

4.0

Unusually broad connector, model, and collaboration surface on paper, but per-feature depth and quality are unverified (README-sourced) and the project is far less proven than open-webui-class tools - so 4, not top, with low confidence.

https://github.com/MODSetter/SurfSense recorded 2026-06-17
feature/connector matrix: 100+ LLMs, 27+ connectors, hybrid RAG, multimodal gen, RBAC collab

Unchanged since 2026-07-30 (last edited, not re-checked)