How it works

Three moves. No magic.

What you do is simple. What happens underneath is engineering — and it is inspectable.

01

Connect

Files, folders, feeds and systems. Start with one PDF; add sources as you go. Parser turns everything into high-fidelity structured evidence.

02

Prove

Verifiable Knowledge adjudicates conflicting evidence into one governed knowledge base — origin, citations, versions — served to any model, agent or person.

03

Stay correct

Continuous monitoring detects change in your sources; new evidence re-enters through Parser and is adjudicated and versioned. Never a direct edit.

Under the hood

The engineering, revealed.

Why it works

Three pillars — the reason this is infrastructure, not another chatbot.

The honest part

What happens when the knowledge base is not sure.

a decision, not an error
When evidence is insufficient, Orves abstains.

Every claim carries a resolution state. Verified means the evidence is there — go read it. Abstained means it is not — and no model gets to fill the gap with fluency.

# the API, telling the truth (pilot)
{ "claim": "renewal_term",
  "state": "ABSTAINED",
  "reason": "insufficient_evidence" }

Build AI systems that know where every answer came from.

Free sandbox — real API, sample corpus, verify a real certificate. Buy credits when it earns it.