Onyx is an open-source AI platform with 32,600+ stars on GitHub that gives teams AI chat and enterprise search over all their workplace knowledge.
Onyx (formerly Danswer) answers the question every company has: how do employees ask an AI about internal knowledge without leaking it to a public chatbot? You self-host Onyx, connect your tools — documents, wikis, tickets, chats — and it indexes them into an agentic RAG system with hybrid search and permission-aware results. On top of that retrieval core sit AI chat that works with every major LLM, deep-research reports, custom agents with their own knowledge subsets and instructions, live web search, and external actions through MCP so agents can complete tasks end to end. Because it is self-hosted, your data stays inside your infrastructure. The project is under very active development, updated in October 2026, and also offered as a managed cloud.
Permission awareness is the detail enterprises care about most: Onyx mirrors the access controls of the systems it indexes, so an employee only ever sees answers drawn from documents they were already allowed to read — the property that separates a deployable product from a demo. Its connector catalog spans the tools companies actually use, from Google Drive and Slack to Confluence, Jira, GitHub and Salesforce, and scheduled syncing keeps the index current. That combination has made it one of the most popular self-hosted answers to the ‘ChatGPT for our company knowledge’ problem.
Onyx’s open-source edition is free to self-host. Onyx Cloud and the Enterprise edition add managed hosting, advanced security and support at paid tiers — see onyx.app for current plans.
Companies and teams that want a ChatGPT-style experience grounded in their own documents and tools, without sending internal knowledge to third-party chatbots.