"author:pbellisan" matching MCP servers:

Enables an AI client to query read-only synthetic LLM usage data for per-model costs, latency, failures, period comparisons, savings breakdowns, provider failure and fallback rates, and slowest analyses.