See AI spend like cloud spend
Track LLM, Bedrock, and Vertex spend per model and per source, with cache-token economics and visibility into the adjacent "iceberg" cost of agentic AI.
AI spend is exploding and invisible. It hides across providers and app features, cache tokens are priced separately, and finance has no idea which model or workload is driving the bill.
Built to do the work, not just show it
Per-model, per-source
Break AI spend down by model and by the app surface that generated it.
Cache-token economics
Capture cache-creation and cache-read tokens, which providers price separately from input.
Coverage tracking
Know which sources are instrumented and where you’re blind.
Agentic "iceberg" cost
Surface adjacent costs of multi-agent workflows, not just the headline model call.
From signal to result
Instrument
Usage is logged per call with model + source + token detail.
Aggregate
Spend rolls up by model, source, and time.
Attribute
Tie AI spend back to the workload driving it.
- Cache tokens are captured and priced (write 1.25× / read 0.10× input) — fixing a real under-report.
- Managed AI usage is attributed at the org level.
- Feeds AI usage billing + Anthropic reconciliation downstream.
Questions
Which providers are covered?
LLM APIs plus Bedrock and Vertex, per model and per source.
Do you count cache tokens?
Yes — cache-creation and cache-read tokens are captured and priced separately, as providers bill them.