Skip to main content
This is the vocabulary the product, the docs, and the website all use. Each concept is labeled by maturity so there’s never ambiguity about what’s available to you today.
How we label maturity. GA = generally available. Early Access = real, but off by default and offered to design partners / on request. Direction = where we’re headed, not a feature you can turn on. We never describe an Early Access or Direction capability as if it ships today.

The idea

A change can be authorized and still be dangerous. Your pipeline confirms the actor is allowed to make a change; it cannot tell you the service being changed is mid-incident, or that the cluster is carrying standing risk right now. ChangeGuard AI makes change the primary object: it evaluates a proposed change against the live state of the environment it will land in, returns a risk decision before the change executes, and — after execution — records what actually followed against the change that caused it.
The decision layer for infrastructure change is our term for the long-term direction: the same judgment applied to production change wherever it originates, not only in Kubernetes. That is direction, not a feature you can turn on. Kubernetes is the wedge and the proving ground, and everything below labeled GA is what ships today.

Generally available today

Change history & attribution — GA

Continuous tracking of every deployment, config, and GitOps change across your fleet, with each failing workload correlated to the change that caused it — including the diff that shipped. This is the foundation the rest builds on.

Readiness score (CSC) — GA

A deterministic, explainable 0–100 readiness score. Same inputs, same score — which is what lets you rely on it as a signal. Every point ties to a concrete factor you can inspect. CSC is the stable identifier used in the API and the dashboard; in prose we call it the readiness score.

Pre-flight check — GA

An on-demand “is it safe to ship right now?” verdict: SHIP / HOLD / BLOCK, with the score and reasons. It is advisory — a recommendation, never an enforced gate, unless you choose to wire it into your own pipeline.

Incident investigation — GA

Detect a failing workload, explain the root cause with cited evidence, propose a fix, and describe how to verify it. Analysis you can check, not a black-box verdict.

Current Understanding — GA (where incident context is enabled)

The maintained, change-centric read of an incident — what someone joining right now needs to know. Available where incident context is enabled for your tenant.

Autonomy model — GA

Four levels — Observe → Advise → Approve → Auto — with Advise as the default. You decide how much ChangeGuard AI may do on its own. See Policy-governed remediation.

Policy-governed remediation — GA, opt-in

Detection → diagnosis → policy-gated execution → verification → rollback. Off until you enable it and grant a narrow, namespaced RBAC Role. Below Auto, every action waits for a human to approve it. ChangeGuard AI never claims a fix succeeded it hasn’t verified against real workload health, and on failed verification it performs exactly one audited rollback, then hands off to a human.

Security scanning — GA, plan-gated

CIS benchmarks (KubeBench), CVEs (Grype), deprecated APIs (Pluto), SBOM (Syft), and runtime signals (Falco). Findings go to the control plane; your images and manifests don’t.

Compliance mapping — GA, plan-gated

Maps scan, RBAC, and SBOM evidence to framework controls (SOC 2, PCI, HIPAA, and others). This is support for evidencing controls — it helps you show your work. It is not a certification, and ChangeGuard AI does not certify you as compliant.

What ChangeGuard AI does not do today

Worth stating plainly, because the labels above only tell you what exists.
  • It does not learn from outcomes. The outcome loop is instrumented — a verdict is recorded before a change, and what followed is recorded after — but no model, threshold, policy or prompt is modified by an outcome. Recall of similar past incidents (Early Access) is retrieval, not retraining.
  • It does not act on its own by default. Autonomy ships at Advise. Execution requires you to raise the level and define a policy.
  • It does not cover infrastructure beyond Kubernetes. Terraform, IAM, cloud resources, networking and databases are direction, not shipping capability.

Early Access

These are real and running, but off by default and offered to design partners or on request. Don’t plan production dependencies on them yet.
If you don’t see one of these in your tenant, that’s expected — they’re not self-serve. Contact support if you’d like to join the Early Access program.

Components

A note on honesty

If a capability isn’t listed as GA above, treat it as not-yet-generally-available and check with your ChangeGuard AI contact before depending on it. The website uses these exact same labels — so what you evaluate is what you can turn on.