Onyx (formerly Danswer) is an open-source AI platform that serves as the application layer for large language models, combining enterprise search, retrieval-augmented generation (RAG), custom agents, and deep research capabilities in a single self-hostable package. With 25,200 GitHub stars, MIT-licensed Community Edition, and backing from Y Combinator, Onyx is used by enterprises including Netflix, Ramp, and Thales Group. The platform provides 50+ built-in indexing connectors for tools like Slack, Confluence, Google Drive, GitHub, and more, with continuous sync to keep knowledge bases current. Core features include agentic RAG with hybrid indexing, multi-step deep research workflows, custom AI agents with scoped knowledge, sandboxed code execution for data analysis, voice mode with TTS/STT, and image generation. Onyx is model-agnostic, supporting self-hosted models via Ollama, LiteLLM, and vLLM alongside commercial APIs from Anthropic, OpenAI, and Google. Two deployment modes are available: Lite mode under 1GB RAM for basic chat, and Standard mode with full infrastructure including vector indexing, job queues, and Redis caching. Enterprise features include SSO (OAuth, OIDC, SAML), RBAC, analytics, audit trails, and white-labeling.