Envoy AI Gateway MCP request duration histogram

Declared as a constant in internal/metrics/mcp_metrics.go with the documented dimension error.type. NO EXPLICIT BUCKET BOUNDARIES are set for the MCP family in that file, so the OTel SDK's DEFAULT explicit-bucket advice applies and the ladder is whatever the SDK version ships - a case where histogram_buckets is genuinely null rather than merely unrecorded, and where the bucket list can change with an SDK bump. Contrast the four gen_ai.* histograms, which all set boundaries explicitly. Recorded with null buckets deliberately: the schema says a histogram with unknown buckets gives quantiles that are wrong rather than absent, so this record is a flag to read the exporter's own bucket line rather than to trust a quantile. Envoy AI Gateway main (envoyproxy/ai-gateway), read 2026-10-05. Metrics go through an OpenTelemetry MeterProvider (internal/metrics/metrics.go NewMeterFromEnv), not a bare prometheus_client registry: a Prometheus reader is ALWAYS attached, and a console or OTLP exporter is added additionally when OTEL_METRICS_EXPORTER is 'console', or when it is neither 'none' nor 'prometheus' and an OTLP endpoint env var is set. OTEL_SDK_DISABLED=true turns off the extra exporters but not the Prometheus reader. That means the metric names are dotted OTel names (gen_ai.server.*), which the Prometheus exporter renders with dots replaced by underscores and a unit suffix appended - so the scraped series is NOT the literal constant in genai.go. Names follow the OpenTelemetry GenAI semantic conventions, not any engine's private scheme.