diff --git a/cmd/adapter/adapter.go b/cmd/adapter/adapter.go
index 13ef3f8d..36a72ec7 100644
--- a/cmd/adapter/adapter.go
+++ b/cmd/adapter/adapter.go
@@ -271,11 +271,23 @@ func (cmd *PrometheusAdapter) addResourceMetricsAPI(promClient prom.Client, stop
 		return err
 	}
 
+	config, err := cmd.Config()
+	if err != nil {
+		return err
+	}
+	config.GenericConfig.EnableMetrics = false
+
 	server, err := cmd.Server()
 	if err != nil {
 		return err
 	}
 
+	metricsHandler, err := mprom.MetricsHandler()
+	if err != nil {
+		return err
+	}
+	server.GenericAPIServer.Handler.NonGoRestfulMux.HandleFunc("/metrics", metricsHandler)
+
 	if err := api.Install(provider, podInformer.Lister(), informer.Core().V1().Nodes().Lister(), server.GenericAPIServer, nil); err != nil {
 		return err
 	}
diff --git a/pkg/client/metrics/metrics.go b/pkg/client/metrics/metrics.go
index b54067f2..0faaed82 100644
--- a/pkg/client/metrics/metrics.go
+++ b/pkg/client/metrics/metrics.go
@@ -18,11 +18,16 @@ package metrics
 
 import (
 	"context"
+	"net/http"
 	"net/url"
 	"time"
 
 	"github.com/prometheus/client_golang/prometheus"
 
+	apimetrics "k8s.io/apiserver/pkg/endpoints/metrics"
+	"k8s.io/component-base/metrics"
+	"k8s.io/component-base/metrics/legacyregistry"
+
 	"sigs.k8s.io/prometheus-adapter/pkg/client"
 )
 
@@ -30,18 +35,29 @@ var (
 	// queryLatency is the total latency of any query going through the
 	// various endpoints (query, range-query, series).  It includes some deserialization
 	// overhead and HTTP overhead.
-	queryLatency = prometheus.NewHistogramVec(
-		prometheus.HistogramOpts{
-			Name:    "cmgateway_prometheus_query_latency_seconds",
-			Help:    "Prometheus client query latency in seconds.  Broken down by target prometheus endpoint and target server",
-			Buckets: prometheus.ExponentialBuckets(0.0001, 2, 10),
+	queryLatency = metrics.NewHistogramVec(
+		&metrics.HistogramOpts{
+			Namespace: "prometheus_adapter",
+			Subsystem: "prometheus_client",
+			Name:      "request_duration_seconds",
+			Help:      "Prometheus client query latency in seconds.  Broken down by target prometheus endpoint and target server",
+			Buckets:   prometheus.DefBuckets,
 		},
-		[]string{"endpoint", "server"},
+		[]string{"path", "server"},
 	)
 )
 
-func init() {
-	prometheus.MustRegister(queryLatency)
+func MetricsHandler() (http.HandlerFunc, error) {
+	registry := metrics.NewKubeRegistry()
+	err := registry.Register(queryLatency)
+	if err != nil {
+		return nil, err
+	}
+	apimetrics.Register()
+	return func(w http.ResponseWriter, req *http.Request) {
+		legacyregistry.Handler().ServeHTTP(w, req)
+		metrics.HandlerFor(registry, metrics.HandlerOpts{}).ServeHTTP(w, req)
+	}, nil
 }
 
 // instrumentedClient is a client.GenericAPIClient which instruments calls to Do,
@@ -63,7 +79,7 @@ func (c *instrumentedGenericClient) Do(ctx context.Context, verb, endpoint strin
 				return
 			}
 		}
-		queryLatency.With(prometheus.Labels{"endpoint": endpoint, "server": c.serverName}).Observe(endTime.Sub(startTime).Seconds())
+		queryLatency.With(prometheus.Labels{"path": endpoint, "server": c.serverName}).Observe(endTime.Sub(startTime).Seconds())
 	}()
 
 	var resp client.APIResponse