Onyx

Onyx is an open source AI platform that serves as the application layer for LLMs, combining RAG, web search, code execution, deep research, and custom agents in a single self-hostable interface. It supports all major LLM providers — both self-hosted (Ollama, LiteLLM, vLLM) and proprietary (OpenAI, Anthropic, Gemini) — and ships with 50+ indexing connectors plus MCP support. Onyx is available in a lightweight "Lite" mode for quick testing and a full "Standard" mode for production teams, with an Enterprise Edition for larger organizations needing SSO, RBAC, and analytics.

• Agentic RAG — Hybrid keyword + vector index combined with AI agents delivers best-in-class search and answer quality.
• Deep Research — Multi-step research flow generates in-depth reports; ranked at the top of the deep research leaderboard as of Feb 2026.
• Custom Agents — Build AI agents with unique instructions, knowledge bases, and external actions via MCP integrations.
• Web Search — Supports Serper, Google PSE, Brave, SearXNG, Firecrawl, Exa, and a built-in web crawler for up-to-date information.
• Code Execution — Run code in a sandbox to analyze data, render graphs, or modify files directly within the chat interface.
• Voice Mode — Chat via text-to-speech and speech-to-text for hands-free interaction.
• Image Generation — Generate images from natural language prompts inside the chat UI.
• Artifacts — Produce downloadable documents, graphics, and other file outputs from agent workflows.
• 50+ Connectors — Index knowledge from Slack, Confluence, Google Drive, Notion, GitHub, and dozens more out of the box.
• Enterprise Features — SSO (Google OAuth, OIDC, SAML), SCIM provisioning, RBAC, usage analytics, query history auditing, and white-labeling.
• Flexible Deployment — Deploy with a single curl command; supports Docker, Kubernetes, Helm/Terraform, and major cloud providers.
• MIT Licensed (CE) — The Community Edition is freely available under the MIT license; self-host with full source access.