Spendassay Meter

Know what the AI in your product costs, per customer.

You ship an AI feature. Meter shows what it costs for every customer, feature, and agent run, so you can see margin per account, catch a customer whose usage is torching your economics, and attribute AI COGS for usage-based billing.

Per-customer costObserve-onlyNever stores prompts
AI cost by customer · last 30 daysDemo data · synthetic
Total AI cost to serve Counted
$41,900
across 128 customers · 18.4M requests
acme-corp · 2.6M req · below your avg cost/req
$4,210
globex · 840K req
$2,980
initech · 22K req · 38× your avg cost/req
$1,910
Priced from cited vendor ratesevery request tagged to a customer
How it works

How to measure AI cost per customer

Point your AI traffic through the proxy or report it with the SDK, tag each request, and Meter does the pricing and the roll-up.

01

Point your traffic

Drop in the spendassay-proxy gateway (one base-URL swap) or the spendassay-meter SDK (report usage from the provider response, with no gateway in your hot path).

02

Tag each request

Attach the customer, feature, agent-run, and environment the call belongs to. That's the whole integration.

03

See cost per customer

Every request is priced from cited vendor rates and rolls up per customer and feature. Add revenue to turn cost into margin.

What you get

The number your vendor bill can't give you.

Cost per customer

The number you can't get from your vendor bill: what the AI in your product costs for each account you serve.

Cost per feature and agent

Which feature or agent loop is eating the model budget. Attribute every token to the thing that spent it.

Anomaly alerts

Get told when an account spikes past its trailing average, before it shows up as a margin surprise.

Margin, when you're ready

Add revenue per customer and cost becomes gross margin, with margin-negative accounts flagged.

Pricing

Priced on metered requests, never on your spend.

Flat, volume-based, with a free tier to start. We meter the traffic; we never take a cut of your inference.

Free
up to 250K metered requests / mo
$0

Enough to wire it in and see cost per customer on real traffic.

Sign up free
  • Per-customer and per-feature cost
  • Proxy or SDK on-ramp
  • 30-day retention
  • 1 ingest key
Most popular
Build
up to 2M metered requests / mo
$99
per month

The full picture for a growing AI product.

Sign up
  • Everything in Free
  • All attribution: customer, feature, agent, environment
  • Cost anomaly alerts
  • Full retention
  • Unlimited ingest keys
Operate
up to 20M metered requests / mo
$499
per month

Attach revenue for per-customer margin, and a weekly digest.

Sign up
  • Everything in Build
  • Revenue import for per-customer margin
  • Weekly digest
  • Higher volume
Enterprise
over 20M metered requests / mo
Custom

Enterprise controls for high-volume AI products.

Talk to sales
  • Everything in Operate
  • SSO / SCIM, single-tenant region
  • SOC 2 readiness report and DPA
  • Priority support

Overage is soft, about $60 per additional 1M metered requests on Free and Build, and about $25 on Operate, never a hard cutoff that blinds you.
Bring your own provider keys. Meter reads token counts and your tags, never prompt or completion bodies.

Posture

Observe-only, by design.

Meter sits in the observability half of the request path. It reads token counts and your tags and discards request and response bodies. It never stores your prompts or completions, and it never routes or throttles your traffic. The proxy is open source, so you can read every line before you run it. See the full security posture.

Spendassay Meter

See what your product's AI costs, per customer.

Wire in the proxy or the SDK, tag your traffic, and watch cost roll up by account. Start free — no credit card.