Skip to main content
Onyx is an open-source AI platform that gives your team a powerful chat interface backed by your organization’s knowledge. It works with any LLM — cloud or self-hosted — and ships with advanced retrieval, agent capabilities, and enterprise-grade security built in. Onyx runs entirely on your infrastructure. You can deploy it in minutes with Docker Compose, or configure it for large-scale production on Kubernetes or Terraform.
Don’t want to self-host? Try Onyx Cloud — a fully managed version with no setup required.

​
Key features

Custom Agents

Build AI Agents with unique instructions, dedicated knowledge bases, and tool access. Agents can act on external systems via MCP.

RAG & hybrid search

Best-in-class hybrid search combined with a knowledge graph for uploaded files and documents ingested from connectors. Stays accurate at tens of millions of documents.

40+ connectors

Pull documents, metadata, and access permissions from Slack, Confluence, GitHub, Google Drive, Jira, Notion, and 40+ more sources.

Deep Research

Multi-step agentic research that synthesizes in-depth answers across your entire knowledge base.

Actions & MCP

Give Agents the ability to interact with external systems using the Model Context Protocol.

Web search

Browse the web with Google PSE, Exa, Serper, an in-house scraper, or Firecrawl — directly from chat.

Code interpreter

Execute code to analyze data, render graphs, and create files without leaving the chat interface.

Image generation

Generate images from natural language prompts within any conversation.

Collaboration

Chat sharing, feedback collection, user management, usage analytics, and role-based access control.

​
LLM compatibility

Onyx works with all major LLM providers through LiteLLM:
  • Cloud providers: OpenAI, Anthropic, Google Gemini, Cohere, and more
  • Self-hosted models: Ollama, vLLM, and any OpenAI-compatible endpoint
  • Configurable per feature: Use different models for chat, search, and embeddings

​
Enterprise-ready

Onyx is built for teams of all sizes, from solo developers to global enterprises:
  • Enterprise search: Custom indexing and retrieval that remains performant at scale — tested to tens of millions of documents
  • Security: SSO via OIDC, SAML, and OAuth2; role-based access control; encrypted credential storage
  • Document permissioning: Mirrors access controls from external apps so users only retrieve documents they can access
  • Management UI: Distinct user roles — basic, curator, and admin — with a full admin panel

​
Licensing

Onyx ships in two editions:
EditionLicenseDescription
Community Edition (CE)MITFully open source. Free to use, modify, and self-host.
Enterprise Edition (EE)CommercialAdds features for larger organizations: advanced SSO, audit logging, multi-tenancy, and more.
The Community Edition covers most use cases. Enterprise Edition features are gated behind a license key and require a paid plan. See onyx.app/pricing for details.

​
Next steps

Quick start

Deploy Onyx in minutes using the one-line install script or Docker Compose.

Architecture

Understand how the backend, vector search, and Celery workers fit together.

Connectors

Connect Onyx to your organization’s knowledge sources.

Build docs developers (and LLMs) love

Get started for freeTalk to us