Onyx
Self-hosted application layer for LLMs with chat, RAG, web search, code execution, and agents.
About
Onyx is an open source application layer for LLMs that teams self-host as a full chat workspace. The interface supports agentic retrieval-augmented generation over connected knowledge, multi-step deep research, custom agents with their own instructions and knowledge, and web search through providers such as Serper, Google PSE, Brave, and SearXNG, plus sandboxed code execution, document generation, and a voice mode. More than 50 indexing connectors ingest sources out of the box, with additional integrations available through the Model Context Protocol. It connects to most LLM backends, from self-hosted Ollama, vLLM, and LiteLLM to OpenAI, Anthropic, and Gemini. Deployments range from Onyx Lite, a lightweight chat and agents build that runs in under 1 GB of memory, to the full stack with vector and keyword indexing, background workers, Redis, and MinIO, installable via Docker, Kubernetes, Helm, or Terraform. The Community Edition is MIT licensed, and an Enterprise Edition serves larger organizations.
Reviews (0)
Leave a Review
No reviews yet. Be the first to review!
Details
- Category
- RAG & Document Retrieval
- Price
- Free
- Platform
- Hybrid
- Difficulty
- Intermediate (3/5)
- License
- MIT
- Added
- May 7, 2026
Related Tools
All-in-one desktop and Docker app for private LLM chat with your documents.
Modular open-source RAG framework for building production document retrieval applications.
Open-source data extraction and indexing engine for RAG applications.
Open-source embedding database for AI applications
Web scraping API that turns websites into clean LLM-ready markdown.
Cloud-native vector database for scalable similarity search
Mentioned in
RAG Is Dead, Long Live RAG: Where Retrieval Is Going
The 'RAG is dead' meme misses what is actually happening. Hybrid retrieval, late-interaction models, agentic...
Max P
Building a Private RAG Stack with Ollama, Qdrant, and AnythingLLM
An end-to-end blueprint for a fully self-hosted RAG system using Ollama for inference, Qdrant for the vector...
Billy C