Onyx is a self-hostable, MIT-licensed knowledge layer that indexes 50+ apps so a team's AI agents retrieve shared organizational context.
Yes, Onyx is open source under the MIT license.
Community Edition free (MIT); Enterprise Edition (SSO, RBAC, analytics, white-labeling) is paid/proprietary (freemium).
Yes, Onyx can be self-hosted (the source is available under the MIT license).
Open-source knowledge layer, formerly published as Danswer, that connects to 50+ applications (docs, tickets, wikis, code) and indexes them into a hybrid search index so a team's AI agents retrieve from one shared, permission-aware context instead of each tool scraping sources independently.
A GTM team running multiple AI agents (a support bot, a sales-research agent, an internal Slack assistant) usually ends up re-indexing the same docs, pricing pages, and tickets in three different tools, each with its own retrieval quality. Onyx's bet is a single owned retrieval layer: index once, let every agent query the same permission-aware index. The honest read: the free MIT tier covers the core retrieval and chat stack, which is most of what a lean team needs, but enterprise-grade access control and white-labeling sit behind a paid tier — worth checking that split against your actual compliance requirements before standardizing on it.