Documentation

From install to first catch.

The first useful outcome is a finding on your own production, in read-only mode, so you can judge its evidence before sudoSRE is allowed to act on anything.

Full installation and integration documentation is shared with design partners during onboarding. This page describes the shape of a deployment.

Three phases.

  1. 01day 1

    Install the cluster agent

    One Helm chart installs the agent, Prod Map, and the investigator into your cluster, pointed at the model endpoint you choose. Everything starts read-only.

    helm install, read-only

  2. 02first day

    Let it learn your production

    Within minutes the Prod Map shows your services and how they connect. Baselines settle over the first hours. Connect Slack, and optionally GitHub or Prometheus for extra detail.

    map in minutes, baselines in hours

  3. 03when ready

    Turn on the first action

    Once you trust the findings, enable rollback for one namespace. Every rollback still waits for a person to approve it in Slack unless your policy says otherwise.

    per action kind, per namespace

Your first deployment changes nothing.

Start read-only. Watch the Prod Map fill in, read the first findings, and check them against what you know about your services. When the evidence holds up, enable rollback for one namespace and approve the first one yourself.

values.yamlillustrative
# values.yaml. Illustrative.
# A first install watches and reports. It changes nothing.
cluster: prod-us-east-1

agent:
  ebpf: true
  prometheus: ""          # optional, read extra metrics

investigator:
  model: claude-opus-5-5  # or your own endpoint
  slackWebhook: existingSecret

actions:
  enabled: false          # turn on later, one kind at a time
  autoApprove: []         # every action needs a person

Put sudoSRE in your cluster.

We are onboarding a small group of Kubernetes teams as design partners. Install takes one Helm chart, starts read-only, and is free during the partnership.