Onyx is an open-source AI platform that provides an application layer for connecting language models to organizational knowledge and tools. It combines a chat and agent interface with retrieval-augmented generation, web search, deep research, code execution, file generation, and integrations for indexing data from many business systems.

Its RAG stack uses hybrid retrieval and agentic search, while custom agents can be configured with their own instructions, knowledge, and actions. Onyx supports web search, MCP-based actions, sandboxed code execution, voice interaction, image generation, and more than 50 indexing connectors. It works with proprietary model providers as well as self-hosted inference systems such as Ollama, vLLM, and LiteLLM.

Onyx can be deployed in a lightweight mode for chat and agents or as a full stack with vector and keyword indexes, background workers, model services, Redis, and object storage for larger organizations. It is suited to teams building private enterprise search, internal assistants, research agents, and collaborative AI workspaces with self-hosting, SSO, and enterprise deployment options.