A new open-source AI platform called Onyx is making the case that enterprises and individual developers shouldn’t have to choose between the convenience of hosted AI services and the control of self-hosted infrastructure — they can have both.
Onyx, which can be installed with a single command and deployed via Docker, Kubernetes, or Terraform, is a feature-rich Chat UI that works with any Large Language Model — from OpenAI’s GPT series and Anthropic’s Claude to open-source models running on local hardware via Ollama or vLLM. And critically, it can run entirely in air-gapped environments with no internet connectivity required.
Built for Teams, Designed for Enterprises
Onyx is the product of onyx-dot-app, a company that has been steadily building the platform as an open-source alternative to proprietary enterprise AI platforms. The Community Edition is freely available under the MIT license, while an Enterprise Edition adds features like SSO (OIDC/SAML/OAuth2), role-based access control (RBAC), document permissioning that mirrors user access from external apps, and usage analytics.
The platform’s feature list is substantial. Onyx ships with:
- Custom Agents: Build AI agents with unique instructions, domain knowledge, and the ability to interact with external systems.
- RAG (Retrieval-Augmented Generation): Hybrid search combining vector and knowledge graph approaches, supporting document ingestion from over 40 external applications.
- Deep Research: An agentic multi-step search capability that autonomously explores topics and synthesizes findings.
- Web Search: Integration with Google PSE, Exa, Serper, Firecrawl, and Onyx’s own in-house scraper.
- MCP (Model Context Protocol): The same protocol Anthropic pioneered, enabling AI agents to interact with external tools and data sources.
- Code Interpreter: Execute code to analyze data, render charts, and generate files directly within conversations.
- Image Generation: Built-in image generation capabilities based on user prompts.
Enterprise Scale, Open Source Freedom
For individual users, Onyx offers Onyx Cloud — a hosted version that requires no deployment. For organizations that need to keep data in-house, the self-hosted option supports deployments of up to tens of millions of documents in the RAG index, with custom indexing and retrieval pipelines that maintain accuracy at scale.
Security is a major differentiator. With SSO and RBAC built in, enterprises can deploy Onyx as a controlled internal tool with fine-grained permissions. Credentials are encrypted, and since everything runs on your own infrastructure, no data leaves your network — a critical requirement for regulated industries like healthcare, finance, and legal.
Connectors to 40+ Data Sources
Onyx’s connector system is one of its most distinctive features. Rather than requiring users to manually upload documents, Onyx can pull knowledge, metadata, and access information directly from a wide range of enterprise applications including Google Workspace, Microsoft 365, Slack, Notion, Confluence, GitHub, and many others. This means an internal AI assistant built on Onyx can answer questions using live data from your organization’s existing tools.
Easy Deployment
Getting started is refreshingly simple. The official install script can be downloaded and run on any supported platform, with Docker and Kubernetes guides available for all major cloud providers including AWS EKS and Azure VMs, as well as Terraform templates for teams that manage infrastructure as code.
The Open Source AI Platform Wars
Onyx enters an increasingly crowded market for open-source AI platforms. But its combination of enterprise-grade features, broad LLM compatibility, and genuine air-gap capability gives it a distinctive position. Unlike platforms that are designed primarily for a single model or provider, Onyx’s model-agnostic architecture means organizations aren’t locked in — they can switch which LLM powers their assistant without changing the interface or workflow.
With its active Discord community, public roadmap on GitHub, and both free and commercial tiers, Onyx represents a compelling option for teams that want the power of modern AI assistants without surrendering control of their data.
The project is available at github.com/onyx-dot-app/onyx and at onyx.app for those who want to try Onyx Cloud first.