Operator quickstart

Deploy the OpenLIT Operator and instrument a sample AI Agent in 5 minutes, with zero code changes.

In this tutorial you’ll deploy an example AI Agent in Kubernetes and automatically capture complete observability — distributed traces, LLM costs, token usage, and agent performance metrics — using the OpenLIT Operator.

Prerequisites

• A Kubernetes cluster with cluster-admin access
• Helm installed
• kubectl configured for your cluster

Don’t have a cluster? You can create one locally with k3d (k3d cluster create openlit-demo) or minikube (minikube start --cpus=2 --memory=4096mb --driver=docker).

1. Install the OpenLIT Platform

   Install the OpenLIT observability platform to collect and visualize traces and metrics.

   Add the Helm repository:

   helm repo add openlit https://openlit.github.io/helm/
   helm repo update
   

   Install the platform:

   helm install openlit openlit/openlit \
     --namespace openlit \
     --create-namespace
   

   Wait for the platform pods to reach Running status:

   kubectl get pods -n openlit
   
2. Install the OpenLIT Operator

   Install the operator to enable zero-code instrumentation:

   helm install openlit-operator openlit/openlit-operator
   

   Verify the operator is running:

   kubectl get pods -n openlit -l app.kubernetes.io/name=openlit-operator
   

   Expected output:

   NAME                                READY   STATUS    RESTARTS   AGE
   openlit-operator-7b9c8d5f7b-xyz12   1/1     Running   0          30s
   
3. Create an AutoInstrumentation resource

   Create an AutoInstrumentation custom resource that defines which pods to instrument and where to send telemetry:

   kubectl apply -f - <<EOF
   apiVersion: openlit.io/v1alpha1
   kind: AutoInstrumentation
   metadata:
     name: quickstart-instrumentation
     namespace: default
   spec:
     selector:
       matchLabels:
         instrumentation: openlit
     python:
       instrumentation:
         enabled: true
     otlp:
       endpoint: "http://openlit.openlit.svc.cluster.local:4318"
   EOF
   

   This tells the operator to inject instrumentation into any pod in the default namespace that has the label instrumentation: openlit.

   Already have AI applications running? Instrumentation is injected at pod startup. Restart existing pods to enable it:

   kubectl rollout restart deployment <your-deployment-name>
   
4. Deploy the example AI Agent

   Deploy a sample AI Agent built with CrewAI. This deployment already has the instrumentation: openlit label, so it will be automatically instrumented:

   kubectl apply -f https://raw.githubusercontent.com/openlit/openlit/main/operator/examples/test-application-deployment.yaml
   

   Check that the pod starts with the instrumentation init container:

   kubectl get pods -l app=openlit-test-app
   kubectl describe pod -l app=openlit-test-app | grep -A5 "Init Containers"
   
5. View traces and metrics in OpenLIT

   Port-forward to the OpenLIT dashboard:

   kubectl port-forward -n openlit svc/openlit 3000:3000
   

   Open http://localhost:3000 in your browser and navigate to the Traces section. You’ll see:

   • Service overview — your openlit-test-app service with health metrics
   • Trace timeline — individual traces for HTTP requests and LLM API calls
   • LLM operations — detailed spans showing token usage per call
   • Cost tracking — estimated costs based on token usage
   • Performance metrics — response times, error rates, and throughput

Next steps

Installation

Full installation guide with prerequisites, upgrade paths, and troubleshooting

Configuration

Configure the operator and AutoInstrumentation resources in detail

Instrumentations

Annotate your own deployments and choose an instrumentation provider

Destinations

Send telemetry to your existing observability stack