Inferly captures metadata about every LLM API call — model, tokens, cost, latency, success — aggregates it, and shows it on a clean dashboard with spend alerting. Never touches prompt content.