# Onyx
Onyx is an [[Open Source]] (MIT-licensed) application layer for [[Large Language Models (LLMs)]]. It provides a [[Self-hosting|self-hostable]] platform that combines AI chat with [[Retrieval-Augmented Generation (RAG)]], web browsing, code execution, and agentic workflows. It works with virtually any LLM provider, whether self-hosted (Ollama, LiteLLM) or proprietary (Anthropic, OpenAI, Gemini).
## Key Features
- **Agentic RAG**: hybrid indexing combined with [[AI Agents]] for information retrieval
- **Deep Research**: multi-step research flows
- **Custom Agents**: user-configurable agents with unique instructions and knowledge bases
- **Web Search Integration**: multiple providers supported (Serper, Brave, SearXNG)
- **Code Execution**: sandboxed environment for data analysis and file manipulation
- **Voice Interaction**: text-to-speech and speech-to-text
- **Artifact Generation**: documents, graphics, and downloadable outputs
- **[[Model Context Protocol (MCP)]]** support
## Tech Stack
Primarily [[Python]] (63%) and [[TypeScript]] (31%), with some Go. Multi-component architecture including backend services, frontend web app, CLI tools, desktop client, and browser extensions.
## Editions
- **Community Edition**: free, core features
- **Enterprise Edition**: SAML/OIDC SSO, RBAC, usage analytics, white-labeling, PII filtering
## References
- https://github.com/onyx-dot-app/onyx
## Related
- [[Large Language Models (LLMs)]]
- [[Retrieval-Augmented Generation (RAG)]]
- [[AI Agents]]
- [[Model Context Protocol (MCP)]]
- [[Self-hosting]]
- [[Open Source]]