Compare
The AI analyst, made checkable. Litlyx was one of the first open-source analytics tools to ship a built-in AI analyst ("Lit"). The starting idea is the same as ours: numbers alone are not enough, someone has to interpret them. Where we differ is how that interpretation gets verified.
| Capability | Litlyx | Vitrus |
|---|---|---|
| Cookie-free, no consent banner | ✓ | ✓ |
| Built-in AI interpretation | ✓ | ✓ |
| Clickable query behind every number | — | ✓ |
| Unprovable AI sentences auto-dropped | — | ✓ |
| Proactive digest (arrives unasked) | reactive chat | proactive digest |
| Local LLM / bring your own key | — | ✓ |
| AI referral ↔ AI crawler split | — | ✓ |
| MCP server for AI agents | — | ✓ |
| Self-hostable | Apache-2.0, MongoDB + Redis | Apache-2.0, embedded |
| Free-form chat interface | ✓ | — |
Features we don't have are shown as “—”. A comparison that hides its own gaps would contradict the one thing this product claims.
If you want a conversational interface to interrogate your data, Lit is more focused on that than we are. Our digest is proactive and template-driven; we have no free-form chat surface at all.
When you need to be able to check a number the AI produced. Every sentence we emit is bound to an evidence bundle, and a sentence containing an unverifiable number is dropped before it ships.
The architecture. Litlyx hands the model your data and shows you the answer. In Vitrus the numbers come from deterministic queries first, the model only sees that bundle, and on the way out every number is matched against the evidence. No match, no sentence.
Probably — but through the same gate. Chat would also have to be incapable of producing an unverifiable number, otherwise the one thing that makes this product different disappears.
Other comparisons: