Raven

Your team's knowledge, on your infrastructure.

A self-hostable, multi-tenant RAG platform with built-in voice, chat, and edge deployment. GDPR-ready out of the box. Bring your own models.

Built for teams who own their infrastructure.

Three properties Raven holds to that most hosted RAG products don't.

  • Self-hostable, by design.

    Run on your own server, your own VPC, or a Raspberry Pi at the edge. Your data never leaves your network. No upsell wall, no hidden telemetry.

  • Multi-tenant from day one.

    Built for teams: workspaces, role-based access, audit trails. SOC 2 and GDPR alignment baked into the schema, not bolted on later.

  • AI that fits your stack.

    Bring your own models — Ollama, OpenAI, Anthropic, Groq, vLLM, anything OpenAI-API-compatible. pgvector + BM25 hybrid search out of the box.

Three surfaces, one knowledge base.

Voice, chat, and ingestion channels — share the same indexes, the same access controls, the same audit trail.

Voice screenshot

Ready to give your team a brain that doesn't leak?

Five minutes from docker compose upto a working voice + chat search over your team's documents.

Read the self-host guide

Free if you run it. Reasonable if we run it.

Self-host the whole thing for free, or let us run it for you on managed Cloud with SLA, SSO, and audit logs.

Self-Hosted

Free forever. Source-available.

₹0

Self-host in 5 min

Cloud

Managed by us, ready to scale.

From ₹1,700 / seat / month

See pricing

Frequently asked questions

    • Is Raven really free to self-host?

      Yes. No telemetry, no upsell wall. The source is public and you keep it.

    • Can it run on a Raspberry Pi?

      Yes. Edge deployment is a first-class target — Raven ships a Compose variant tuned for ARM64 + low memory.

    • What is the difference between Self-Hosted and Cloud?

      Identical software. Cloud is run by us with SLA, SSO, and audit logs in the region of your choice.

    • Where does my data go?

      Self-hosted: nowhere it didn't already go. Cloud: only to our infrastructure, in the region you pick.

    • Which models does it support?

      Anything OpenAI-API-compatible — Ollama, OpenAI, Anthropic, Groq, vLLM. Embeddings via pgvector with BM25 hybrid search.

    • How is this different from a hosted Notion AI or a wrapper?

      You own the data, the schema, and the keys. Raven is the platform; the surface is yours to extend.