Quivr

Quivr is an open-source RAG platform (~38k GitHub stars, Apache-2.0, Y Combinator-backed) that popularized the 'second brain' concept: dump in your documents and chat with them. Its pitch was opinionated RAG, any file type, any LLM (OpenAI, Anthropic, Mistral, Gemma, local models via Ollama), and custom parsers, distributed as a Python package (pip install quivr-core) that developers embed in their own apps, with Megaparse integration for ingestion.

This entry carries a staleness flag. The open-source repo is not formally archived but shows a stalled pattern: last meaningful commits in mid-2025, no release since early 2025. The company's commercial focus has visibly shifted to a separate, unrelated customer-support-agent SaaS at quivr.com, and the original product site is a dead domain. Current docs frame it as a lightweight RAG library, not the turnkey full-stack app it once was.

It is included for its historical significance in the RAG wave. For new projects, prefer actively maintained alternatives like Haystack, LlamaIndex, or LangChain; use quivr-core only if you specifically want its opinionated minimal RAG and accept the maintenance risk.

pip install quivr-core and embed the RAG pipeline in your own application (self-hosted).