Self-hosted RAG and knowledge base apps for small businesses
Short answer
We track 7 self-hostable knowledge-base / RAG apps: RAGFlow, AnythingLLM, PrivateGPT, Open Notebook, Onyx (formerly Danswer), FastGPT, MaxKB.
Most popular on GitHub: RAGFlow (92.6k stars).
Lowest official hardware: AnythingLLM (minimum 2 vCPU · 2 GB RAM · 5 GB disk, from €19.99/month on our reference server).
Permissive license (MIT/Apache-2.0, fewest conditions): RAGFlow, AnythingLLM, PrivateGPT, Open Notebook.
No official hardware figure published: PrivateGPT.
Last updated · generated from the same verified data as each app page
Comparison table
| App | What it is | License | Official hardware | Est. server cost | GitHub stars | Status |
|---|---|---|---|---|---|---|
| RAGFlow | RAG engine with deep document parsing (OCR, layout and tables) and cited answers. | Apache-2.0 (Permissive) | recommended 4 vCPU · 16 GB RAM · 50 GB disk (no separate minimum published) | from €69.99/mo | 92.6k | Active |
| AnythingLLM | All-in-one private chat with your documents, agents and an embeddable widget. | MIT (Permissive) | minimum 2 vCPU · 2 GB RAM · 5 GB disk | from €19.99/mo | 67.4k | Active |
| PrivateGPT | API-first private document Q&A that sits on top of your own LLM server. | Apache-2.0 (Permissive) | Not published | Not estimated (no official RAM figure) | 57.6k | Active |
| Open Notebook | Private, self-hosted alternative to Google NotebookLM with your choice of AI models. | MIT (Permissive) | minimum 4 GB RAM | from €19.99/mo | 40.5k | Active |
| Onyx (formerly Danswer) | AI chat and enterprise search connected to 50+ workplace apps. | MIT (Community Edition) + Onyx Enterprise License (ee directories) (Open core) | minimum 4 vCPU · 10 GB RAM · 32 GB disk (recommended 8 vCPU · 16 GB RAM · 500 GB disk) | from €69.99/mo | 32.8k | Active |
| FastGPT | Knowledge-base Q&A and visual AI workflow platform. | FastGPT Open Source License (Apache-2.0 with conditions) (Custom / conditional) | minimum 2 vCPU · 4 GB RAM (recommended 2 vCPU · 8 GB RAM) | from €19.99/mo | 29.9k | Active |
| MaxKB | Enterprise knowledge-base and agent platform with RAG and workflows. | GPL-3.0 (Copyleft (GPL)) | minimum 4 vCPU · 8 GB RAM · 100 GB disk | from €35.99/mo | 23.0k | Active |
Apps from our "Knowledge base & RAG" category, sorted by GitHub stars. We do not rank by quality; pick by the criterion that matters to you. Cost = cheapest Hetzner Cloud Regular Performance (shared AMD vCPU) plan that meets the project's official RAM and vCPU figure, list price read on 2026-10-22, excl. VAT, server only (no LLM API usage, backups or your time).
Frequently asked questions
What is the best self-hosted RAG app for a small business?
Most popular on GitHub: RAGFlow (92.6k stars). Lowest official hardware: AnythingLLM (minimum 2 vCPU · 2 GB RAM · 5 GB disk, from €19.99/month on our reference server). Permissive license (MIT/Apache-2.0, fewest conditions): RAGFlow, AnythingLLM, PrivateGPT, Open Notebook. No official hardware figure published: PrivateGPT. The right choice depends on which of these matters most to you.
Which self-hosted knowledge base needs the least hardware?
AnythingLLM: minimum 2 vCPU · 2 GB RAM · 5 GB disk (from €19.99/month); Open Notebook: minimum 4 GB RAM (from €19.99/month); FastGPT: minimum 2 vCPU · 4 GB RAM (recommended 2 vCPU · 8 GB RAM) (from €19.99/month); MaxKB: minimum 4 vCPU · 8 GB RAM · 100 GB disk (from €35.99/month); RAGFlow: recommended 4 vCPU · 16 GB RAM · 50 GB disk (no separate minimum published) (from €69.99/month); Onyx (formerly Danswer): minimum 4 vCPU · 10 GB RAM · 32 GB disk (recommended 8 vCPU · 16 GB RAM · 500 GB disk) (from €69.99/month).
Which self-hosted RAG apps have an MIT or Apache license?
RAGFlow, AnythingLLM, PrivateGPT, Open Notebook. Onyx (formerly Danswer) uses MIT (Community Edition) + Onyx Enterprise License (ee directories). FastGPT uses FastGPT Open Source License (Apache-2.0 with conditions). MaxKB uses GPL-3.0.
Related
Every figure on this page comes from the linked app pages, which cite the project's own repository and docs. Spotted something wrong? Tell us.