AI Spend Auditor combines provider-native OpenAI and Anthropic organization reports into a read-only spend review without routing model calls or collecting content.
It reads allowlisted cost and aggregate usage endpoints, preserves each provider's native cost units and coverage exclusions, compares equal periods, ranks arithmetic cost contributors, and builds an explicit actual-to-budget brief. A missing provider remains unavailable rather than silently becoming zero. Prompt, response, trace, message, and user-email content are never requested.
What you can do
- summarize_costs — reconcile native cost and usage buckets across connected providers
- compare_periods — calculate absolute and percentage deltas with zero-baseline semantics
- find_cost_drivers — rank arithmetic contributors by model, project, workspace, line item, or tier
- build_budget_brief — compare actual daily pace and a 30-day projection with caller budgets
Who it's for
AI platform leads, engineering managers, finance owners, agencies, and founders operating OpenAI and Anthropic accounts and reconciling spend every day or week.
How to use it
- Connect one or both provider organization admin reporting keys
- Query no more than 180 days with daily buckets
- Preserve provider totals, raw units, exclusions, and partial status independently
- Compare equal-duration windows before investigating a movement
- Use driver and budget outputs as arithmetic evidence, not a diagnosis or automatic control
Getting started
Start with a seven-day summary for both providers. Anthropic documents that Priority Tier cost is excluded and that account coverage differs, so its totals are never presented as complete when those gaps may apply.
