Attribute’s eBPF sensor captures Claude API calls at runtime, extracting model identifiers, token counts, and request metadata. Spend is attributed to the calling workload and downstream customer automatically.
WHAT YOU GET
→ Per-model cost breakdown: Claude Opus, Sonnet, Haiku
→ Token-level cost attribution by calling workload
→ Cost anomaly detection on Anthropic and Claude spend
→ Token economics per customer, feature, or custom unit.
BEST FOR
SaaS teams calling Claude via the Anthropic API directly who need to understand which customers or features drive AI spend.