Product

Four surfaces that turn a bill into an explanation.

ACLP is not a replacement for your provider dashboard. It answers the question the dashboard leaves open: which part of the application produced this spend, and is that the cost of growth or the cost of waste.

  • 01

    Cost & attribution

    Where did the money go?

    Spend attributed by feature, user, model, provider, environment, and session, with each dimension reconciling to the same total. Rows open a filtered view of the requests behind them.

  • 02

    Trends

    What changed, and when?

    Spend, request volume, token composition, average cost per request, and model mix over a selected period, compared with the equivalent period before it.

  • 03

    Usage explorer

    Which request exactly?

    A searchable request ledger with token counts, cost, latency, status, and retry relationships. Prompt and completion text is not part of the record.

  • 04

    Leak detection

    What should we fix?

    Findings for duplicate prompts, retry loops, context growth, oversized prompts, spend spikes, and unexpected model usage, each with evidence and a next step.

Reading a finding

Every number says what kind of number it is.

Observed
Measured directly from the telemetry in the selected period: token counts, request counts, latency, spend.
Estimated
A modelled figure with its method and assumptions shown next to it. Labelled as an estimate wherever it appears.
Suggested
A recommended next step. Suggestions are never applied automatically and never presented as a completed saving.