Input
Parse the request rows
The auditor accepts JSON arrays, records arrays, or NDJSON from common gateway exports.
Paste one trace and see which team, service, and model caused the spend.
Use JSON, a records array, or NDJSON with cost, model, and owner fields. Nothing is uploaded.
Public one-off analysis runs locally in this browser tab.
Results stay empty until you run analysis on pasted JSON or NDJSON with cost plus model, team, tenant, project, or customer fields.
Analyze and Clear are disabled until input is present.
One-off analysis is public. Saved history, shared reports, and monthly review are handled through early access.
Use this page when the bill moved and you need a first owner-level answer before starting an integration.
Input
The auditor accepts JSON arrays, records arrays, or NDJSON from common gateway exports.
Attribution
Team, tenant, customer, project, service, and model fields become the attribution table.
Next step
Copy or download the derived table, then join the waitlist when recurring reports matter.
AI spend is invisible across teams. Engineering sees the bill go up, Finance can't tell which tenant or workload caused it, and chargebacks stall because no one trusts the numbers. This free auditor turns a raw gateway-trace export into the cost-attribution evidence both sides agree on — no vendor demo, no pilot, no data leaving your browser.
Paste a JSON or NDJSON gateway trace from OpenAI, Anthropic, Bedrock, LiteLLM, or your in-house proxy.
See the cost broken down by team, tenant, and model — with an explicit unattributed bucket.
Export attribution as JSON / CSV to attach to the chargeback ticket or RCA.
Pro keeps trace history, runs scheduled diffs, and gives Finance and Engineering one shared chargeback view.
See Pro plans