diff --git a/.github/workflows/verify-k8s.yml b/.github/workflows/verify-k8s.yml
index 9bacc1038..84e6a391f 100644
--- a/.github/workflows/verify-k8s.yml
+++ b/.github/workflows/verify-k8s.yml
@@ -140,6 +140,17 @@ jobs:
           sed -i -e 's,--ca=googleca,--ca=ephemeralca,g' ${{ github.workspace }}/config/deployment.yaml
           # Drop the ct-log flag's value to elide CT-log uploads.
           sed -i -E 's,"--ct-log-url=[^"]+","--ct-log-url=",g' ${{ github.workspace }}/config/deployment.yaml
+          # Switch to one replica to make it easier to test the scraping of
+          # metrics since we know all the requests then go to the same server.
+          sed -i -E 's,replicas: 3,replicas: 1,g' ${{ github.workspace }}/config/deployment.yaml
+          # Expose the prometheus port as a service so tests can grab it
+          # without hitting the k8s API
+          cat <<EOF >> ${{ github.workspace }}/config/deployment.yaml
+              - name: prometheus
+                protocol: TCP
+                port: 2112
+                targetPort: 2112
+          EOF
 
           # From: https://banzaicloud.com/blog/kubernetes-oidc/
           # To be able to fetch the public keys and validate the JWT tokens against
@@ -225,6 +236,25 @@ jobs:
 
           kubectl wait --for=condition=Complete --timeout=90s job/check-oidc
 
+      - name: Validate prometheus metrics exported and look correct
+        run: |
+          cat <<EOF | ko apply -f -
+          apiVersion: batch/v1
+          kind: Job
+          metadata:
+            name: check-prometheus-metrics
+          spec:
+            template:
+              spec:
+                restartPolicy: Never
+                automountServiceAccountToken: false
+                containers:
+                - name: check-metrics
+                  image: ko://github.com/sigstore/fulcio/test/prometheus/
+          EOF
+
+          kubectl wait --for=condition=Complete --timeout=90s job/check-prometheus-metrics
+
       - name: Collect logs
         if: ${{ always() }}
         run: |
diff --git a/README.md b/README.md
index fb9a188a9..4b696957d 100644
--- a/README.md
+++ b/README.md
@@ -41,7 +41,7 @@ There are other targets available in the [`Makefile`](Makefile), check it out.
 
 ## API
 
-The API is defined via OpenAPI, defined [here](openapi.yaml).
+The API is defined [here](./pkg/api/client.go).
 
 ## Transparency
 
diff --git a/cmd/app/serve.go b/cmd/app/serve.go
index 1477eb692..4c1503b05 100644
--- a/cmd/app/serve.go
+++ b/cmd/app/serve.go
@@ -100,19 +100,23 @@ var serveCmd = &cobra.Command{
 		}
 
 		decorateHandler := func(h http.Handler) http.Handler {
-			return http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
-				ctx := r.Context()
-
-				// For each request, infuse context with our snapshot of the FulcioConfig.
-				// TODO(mattmoor): Consider periodically (every minute?) refreshing the ConfigMap
-				// from disk, so that we don't need to cycle pods to pick up config updates.
-				// Alternately we could take advantage of Knative's configmap watcher.
-				ctx = config.With(ctx, cfg)
-				ctx = api.WithCA(ctx, baseca)
-				ctx = api.WithCTLogURL(ctx, viper.GetString("ct-log-url"))
-
-				h.ServeHTTP(rw, r.WithContext(ctx))
-			})
+			// Wrap the inner func with instrumentation to get latencies
+			// that get partitioned by 'code' and 'method'.
+			return promhttp.InstrumentHandlerDuration(
+				api.MetricLatency,
+				http.HandlerFunc(func(rw http.ResponseWriter, r *http.Request) {
+					ctx := r.Context()
+
+					// For each request, infuse context with our snapshot of the FulcioConfig.
+					// TODO(mattmoor): Consider periodically (every minute?) refreshing the ConfigMap
+					// from disk, so that we don't need to cycle pods to pick up config updates.
+					// Alternately we could take advantage of Knative's configmap watcher.
+					ctx = config.With(ctx, cfg)
+					ctx = api.WithCA(ctx, baseca)
+					ctx = api.WithCTLogURL(ctx, viper.GetString("ct-log-url"))
+
+					h.ServeHTTP(rw, r.WithContext(ctx))
+				}))
 		}
 
 		prom := http.Server{
diff --git a/config/deployment.yaml b/config/deployment.yaml
index ac5ddebdb..4161b1473 100644
--- a/config/deployment.yaml
+++ b/config/deployment.yaml
@@ -85,6 +85,7 @@ spec:
     app: fulcio-server
   type: LoadBalancer
   ports:
-    - protocol: TCP
+    - name: http
+      protocol: TCP
       port: 80
       targetPort: 5555
diff --git a/go.mod b/go.mod
index d9ac97b7c..3d2306eb4 100644
--- a/go.mod
+++ b/go.mod
@@ -15,7 +15,8 @@ require (
 	github.com/mitchellh/mapstructure v1.4.3 // indirect
 	github.com/pkg/errors v0.9.1
 	github.com/prometheus/client_golang v1.11.0
-	github.com/prometheus/common v0.29.0 // indirect
+	github.com/prometheus/client_model v0.2.0
+	github.com/prometheus/common v0.29.0
 	github.com/prometheus/procfs v0.7.0 // indirect
 	github.com/sigstore/sigstore v1.0.1
 	github.com/spf13/cobra v1.2.1
diff --git a/pkg/api/client.go b/pkg/api/client.go
index 7f3225dc0..a8d367fcd 100644
--- a/pkg/api/client.go
+++ b/pkg/api/client.go
@@ -38,7 +38,7 @@ const SigstorePublicServerURL = "https://fulcio.sigstore.dev"
 
 // Client is the interface for accessing the Fulcio API.
 type Client interface {
-	// SigningCert sends the provided CertificateRequest to the /api/v1/singingCert
+	// SigningCert sends the provided CertificateRequest to the /api/v1/signingCert
 	// endpoint of a Fulcio API, authenticated with the provided bearer token.
 	SigningCert(cr CertificateRequest, token string) (*CertificateResponse, error)
 }
diff --git a/pkg/api/metrics.go b/pkg/api/metrics.go
index 7b2258ac3..a14a765f5 100644
--- a/pkg/api/metrics.go
+++ b/pkg/api/metrics.go
@@ -29,5 +29,5 @@ var (
 	MetricLatency = promauto.NewHistogramVec(prometheus.HistogramOpts{
 		Name: "fulcio_api_latency",
 		Help: "API Latency on calls",
-	}, []string{"path", "code"})
+	}, []string{"code", "method"})
 )
diff --git a/test/prometheus/main.go b/test/prometheus/main.go
new file mode 100644
index 000000000..f176210fc
--- /dev/null
+++ b/test/prometheus/main.go
@@ -0,0 +1,116 @@
+// Copyright 2021 The Sigstore Authors.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+
+package main
+
+import (
+	"flag"
+	"fmt"
+	"log"
+	"net/http"
+
+	dto "github.com/prometheus/client_model/go"
+	"github.com/prometheus/common/expfmt"
+)
+
+const (
+	latencyMetric = "fulcio_api_latency"
+	certMetric    = "fulcio_new_certs"
+)
+
+func parseMF(url string) (map[string]*dto.MetricFamily, error) {
+	resp, err := http.Get(url) // nolint
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	var parser expfmt.TextParser
+	return parser.TextToMetricFamilies(resp.Body)
+}
+
+func main() {
+	f := flag.String("url", "http://fulcio-server.fulcio-dev.svc:2112/metrics", "set url to fetch metrics from")
+	flag.Parse()
+
+	mf, err := parseMF(*f)
+	if err != nil {
+		log.Fatalf("Failed to fetch/parse metrics: %v", err)
+	}
+
+	// Just grab the api_latency metric, make sure it's a histogram
+	// and just make sure there is at least one 200, and no errors there.
+	latency, ok := mf[latencyMetric]
+	if !ok || latency == nil {
+		log.Fatal("Did not get fulcio_api_latency metric")
+	}
+	if err := checkLatency(latency); err != nil {
+		log.Fatalf("fulcio_api_latency metric failed: %s", err)
+	}
+
+	// Then make sure the cert counter went up.
+	certCount, ok := mf[certMetric]
+	if !ok || certCount == nil {
+		log.Fatal("Did not get fulcio_new_certs metric")
+	}
+	if err := checkCertCount(certCount); err != nil {
+		log.Fatalf("fulcio_new_certs metric failed: %s", err)
+	}
+}
+
+// Make sure latency is a Histogram, and it has a POST with a 201.
+func checkLatency(latency *dto.MetricFamily) error {
+	if *latency.Type != *dto.MetricType_HISTOGRAM.Enum() {
+		return fmt.Errorf("Wrong type, wanted %+v, got: %+v", dto.MetricType_HISTOGRAM.Enum(), latency.Type)
+	}
+	if len(latency.Metric) != 1 {
+		return fmt.Errorf("Got multiple entries, or none for metric, wanted one, got: %+v", latency.Metric)
+	}
+	// Make sure there's a 'post' and it's a 201.
+	var code string
+	var method string
+	for _, value := range latency.Metric[0].Label {
+		if *value.Name == "code" {
+			code = *value.Value
+		}
+		if *value.Name == "method" {
+			method = *value.Value
+		}
+	}
+	if code != "201" {
+		return fmt.Errorf("unexpected code, wanted 201, got %s", code)
+	}
+	if method != "post" {
+		return fmt.Errorf("unexpected method, wanted post, got %s", method)
+	}
+
+	if *latency.Metric[0].Histogram.SampleCount != 1 {
+		return fmt.Errorf("Unexpected samplecount, wanted 1, got %d", *latency.Metric[0].Histogram.SampleCount)
+	}
+	return nil
+}
+
+func checkCertCount(certCount *dto.MetricFamily) error {
+	if *certCount.Type != *dto.MetricType_COUNTER.Enum() {
+		return fmt.Errorf("Wrong type, wanted %+v, got: %+v", dto.MetricType_COUNTER.Enum(), certCount.Type)
+	}
+	if len(certCount.Metric) != 1 {
+		return fmt.Errorf("Got multiple entries, or none for metric, wanted one, got: %+v", certCount.Metric)
+	}
+	if *certCount.Metric[0].Counter.Value < 1 {
+		return fmt.Errorf("Got incorrect cert count, wanted one, got: %f", *certCount.Metric[0].Counter.Value)
+	}
+	return nil
+}