Knowledge platforms shouldn't be data brokers.

Raven exists because most “AI for your team” products are wrappers that ship your documents to someone else's servers, lock you into one model vendor, and treat self-hosting as a nice-to-have at the enterprise tier. We think it should be the default.

Raven is a self-hostable, multi-tenant RAG platform with first-class voice and chat surfaces. It runs on your hardware, with your models, under your access controls — and it does that on a Raspberry Pi just as well as it does on a fleet of VMs.

Who's behind it

Raven is maintained by Jobin Lawrance — previously building Keycloak Service Provider Interfaces, now full-time on auth and platform infrastructure. Based in Bengaluru, India.

The codebase is open on GitHub. Contributions, issues, and pattern discussions welcome.

Source-available under the Raven licence; see LICENSE and ee-LICENSE.