LiyaEngine
Platform

See exactly what
AI is costing you.

Cost tracking runs automatically inside the engine on every call. Spend by model, by domain, by intent — plus a projected end-of-month bill. No black-box invoices, no surprise overages.

Cost Explorer
$696.70
SPENT THIS MONTH
$1,180.40
PROJECTED TOTAL
$900.00
INCLUDED QUOTA
SPEND BY MODEL
claude-sonnet-5$412.20
gpt-4.1$188.40
gemini-2.5$96.10
3
Ways to break down spend — model, domain, intent
6-digit
Cost precision, down to the sub-cent
Live
Per-request cost logged on every call
1
Projected bill, visible before month-end
Cost visibility

No black-box invoices

01

Spend by Model

See exactly what you're paying per provider and per model — claude-sonnet-5, gpt-4.1, gemini-2.5 — broken out side by side, down to the sub-cent.

02

Spend by Domain

Every custom domain logs its own usage. Compare what your support domain costs against your hiring or compliance domain.

03

Spend by Intent

Drill into the specific intent driving the spend — not just "the model," but which exact call is expensive and why.

04

Projected End-of-Month Bill

A live projection of your total spend, including estimated overage against your included request quota — updated with every call.

05

Real Per-Request Cost

Every request logs its real serving model and its real cost — not an estimate, not a flat rate per call.

06

No Hidden Markup

See exactly what each model actually costs. What you're billed is what the provider charged, logged transparently.

Breakdowns

Three ways to slice your spend

Every request is logged once and sliced three ways — by model, by domain, and by intent — so you can find the answer without exporting a spreadsheet.

By Model
Provider comparison

Compare spend across every provider and model you've configured — see which one is actually driving your bill.

6-digit
precision
By Domain
Domain accounting

Every custom domain logs its own usage row. See what support costs against hiring, compliance, or any domain you run.

4-digit
precision
By Intent
Root-cause analysis

Drill past the domain level into the exact intent driving spend — the specific call, not just the general area.

Per-request
precision
Walkthrough

A team catches a runaway spend before the invoice arrives

Mid-month, one intent starts driving 3x its usual spend after a prompt change increased average output length. The projected end-of-month bill jumps visibly on the dashboard — not two weeks later on an invoice.

Drilling into spend by intent, the team finds the exact call responsible, reverts the change, and confirms the projection drops back to baseline the same day.

Projected total · trending
+3.1x
SPIKE ON draft_response
Same day
CAUGHT & REVERTED
draft_response — before$91.20
draft_response — after spike$284.10
FAQ

Common questions

Real. Every request logs its actual serving model and the real cost charged by the provider — not a flat per-call estimate. What you're billed is what the provider charged, logged transparently with no hidden markup.

The projected end-of-month bill updates with every call, so you can see an overage coming before it happens rather than discovering it on the invoice. Budget alerts on Enterprise plans notify your team automatically.

Three levels: by model (which provider/model is driving spend), by domain (which custom domain — support, hiring, compliance, or your own), and by intent (the exact call, down to a single endpoint).

No. Cost logging happens inside the engine alongside the model call itself, not as a separate blocking step — there's no extra round trip before your response comes back.