Onboarding is

We build your full intelligence layer on your existing database in days, not months. No migration, no data team, no setup fee. Billing starts only after the system is live and proven on your data.

Pricing below ↓

The principle

You pay only for the orchestration.

You never pay us for compute. Your LLM and GPU costs are paid by you, directly, to your own cloud or on-prem — we never resell compute or mark it up. We charge only for orchestrating the optimal path through the models, your data, and (where needed) a human. Predictable and per-use — never a per-seat license, never a black-box outcome fee.

Usage-based

Three ways you're billed.

01

Ask

per question

~$0.03

Anyone asks in plain language and gets a governed, audit-traceable answer in seconds. Billed per question answered, not per user — the unit is the account, not the person.

02

Upload & Verify

per document

≥ 2×a question

Upload an Excel or document; it's materialized, verified by independent multi-agent checks, and made queryable. Roughly double a question because each upload triggers verification first.

03

Agents

per scheduled run

~$3≈ 10× a question

Build agents yourself — no build fee, ever — that run on a schedule against your data. Billed only when an agent actually runs. Set it once; pay per execution.

No surprises

What you never pay us for.

  • Your LLM / API costs — paid by you, directly
  • Your GPU / cloud infrastructure — paid by you, directly
  • Per-seat licenses — there are none
  • Onboarding or setup — free
  • “Pay if it works” outcome fees — we don't do outcome pricing

Illustrative

What a mid-size firm spends.

~1,500 people, fully adopted.

Ask~180,000 questions / mo~$5,400
Agents~600 runs / mo~$1,800
Total recurringplus per-document upload-verify~$7,200 / mo

≈ $86K / yr. Illustrative — modeled at full-maturity adoption, not a quote. Cost grows with usage at the same predictable per-question price — never with price hikes.

The logic

Why this model.

Predictable — you always know what a question costs. Fair — you pay for use, not seats. Aligned — your costs scale with the value you pull from your own data, never with how many people log in. Because you own your own compute, the price you pay us is pure orchestration — the part that's genuinely hard to build.