Skip to main content

Quick Start

Get your first AI-powered investigation running in about 10 minutes.

Prerequisites​

  • A Kubernetes cluster (EKS, GKE, AKS, or self-hosted, version 1.16 or later).
  • Helm 3.8 or later installed on your machine (required for OCI charts).
  • kubectl access to your cluster with permission to create namespaces and RBAC.
  • A monitoring system with webhook support (Prometheus AlertManager, Grafana Alerting, or Datadog).
  • A Slack workspace (recommended, for receiving investigation results).

Steps Overview​

1. Create Your Account​

  1. Go to app.opsworker.ai and click Sign Up.
  2. Sign up with Google sign-in or email and password.
  3. Verify your email address.

2. Create a Workspace​

  1. After signup, your organization has a default workspace. You can create additional workspaces at any time.
  2. Name a workspace something meaningful, for example your team name.
  3. Workspaces group your clusters within your organization.

3. Add Your First Cluster​

  1. Navigate to Configurations, then Add Cluster.
  2. Choose the workspace the cluster belongs to.
  3. Optionally enable the Grafana MCP integration inline (you can add integrations later).
  4. Copy the generated Helm install command. It includes your unique cluster token and API key, both filled in for you.

4. Install the Kubernetes Agent​

Run the command from Step 3 in your terminal. It uses an OCI chart, so no helm repo add step is needed:

helm install opsworker-agent oci://public.ecr.aws/k1v3l8a8/opsworker-agent \
  --set clusterToken=YOUR_CLUSTER_TOKEN \
  --set apiKey=YOUR_API_KEY \
  --version=v0.9.3 \
  --timeout=90s \
  --namespace opsworker-agent \
  --create-namespace

The portal pre-fills clusterToken and apiKey, so copy the command directly from the portal rather than typing it by hand.

Verify the agent is running:

kubectl get pods -n opsworker-agent -l app.kubernetes.io/name=opsworker-agent

A default install also creates two bundled MCP server pods in the same namespace. The portal verifies connectivity automatically by polling the agent's status; once it connects, the cluster shows as connected (status Success).

5. Connect Your Alerting System​

Configure your monitoring system to send alerts to OpsWorker via webhook. When you install an alerting integration in the portal, it generates the exact configuration, including the required authorization credentials.

  • Prometheus AlertManager: add a webhook receiver (detailed guide).
  • Grafana Alerting: create a Contact Point of type Webhook (detailed guide).
  • Datadog: add a webhook with the required Authorization header (detailed guide).

6. Connect Slack​

  1. In the portal, go to Integrations, then Slack, and click Connect to Slack.
  2. Authorize OpsWorker in your Slack workspace.
  3. Add namespace routing to map Kubernetes namespaces to the Slack channels where results should be posted.

7. Run Your First Investigation​

You have two options:

  • Wait for a real alert: the next alert from your monitoring system that matches an alert rule on this cluster triggers an investigation automatically.
  • Use AI Chat: open the Chat interface in the portal and ask a question about your cluster.

8. Configure Alert Rules​

By default, configure at least one alert rule per cluster so that matching alerts trigger investigations.

Expected Outcome​

When an investigation completes, you receive a Slack message containing:

  • Alert summary: what triggered the investigation.
  • Root cause analysis: the underlying issue identified by AI.
  • Affected resources: the Kubernetes resources involved (pods, services, deployments).
  • Recommended actions: specific steps to resolve the issue, including kubectl commands.
  • Feedback buttons: rate the investigation quality to improve future results.

The full investigation is also available in the portal under Investigations, where you can see the topology map, collected data, and chat with the investigation for follow-up questions.

Next Steps​