Onyx began as the open-source project Danswer and rebranded as it scaled into a commercial enterprise AI assistant. The product solves the enterprise knowledge problem: information is scattered across docs, wikis, chat tools, drives, and ticketing systems, and employees waste time hunting for answers. Onyx connects to those sources, indexes them with hybrid retrieval-augmented generation, and lets employees ask questions in natural language and get grounded, cited answers that respect each user's access permissions.
Beyond basic search, Onyx offers a collaborative chat experience, deep research workflows, code interpretation, and web search, positioning itself as an open alternative to closed enterprise assistants. Its plug-and-play connectors sync updates from workplace apps in real time while enforcing the same permission structures as the source systems, so sensitive content is not exposed to users who shouldn't see it. The company publishes benchmarks claiming high win rates against general assistants like ChatGPT and Claude on real workplace queries.
The open-source foundation is central to Onyx's appeal. The project has surpassed tens of thousands of GitHub stars, and organizations can self-host it on their own infrastructure for full transparency and data control, with SOC 2 Type II and GDPR considerations addressed for enterprise buyers. This makes it attractive to security-conscious teams that are wary of sending internal knowledge to third-party clouds.
Onyx raised a $10 million seed round in March 2025, co-led by Khosla Ventures and First Round Capital, with Y Combinator and notable angels including operators from Coinbase, Dropbox, and Datadog participating. The company reports enterprise deployments at scale, including Netflix rolling Onyx out across its workforce, and Thales Group among its customers.
Founded by Yuhong Sun and Chris Weaver, Onyx targets engineering, support, operations, and knowledge teams that want a powerful, self-hostable AI assistant grounded in their own data. By pairing open source with enterprise-grade controls, it offers a path to internal AI search without ceding ownership of the underlying knowledge.