AUTOPILOT

Finds what is breaking. Fixes it inside your rules.

Autopilot watches your Kubernetes clusters, servers and databases, explains what is wrong in plain words, and fixes it with your approval or on its own. Every change is checked against your rules, verified and recorded.

Findings on Free · fixes on Pro · autonomous on Business

3:07 a.m. Checkout is restarting.

Usually someone gets paged, opens three tools and guesses at a change.

How it works

You set the rules. It does the work.

01

Choose what it watches

Watch a Kubernetes or OpenShift cluster, a server or a database.

Example data.

02

Choose how far it goes

Set autonomy for each scope: off, shows only, asks first, or acts on its own.

Example data.

03

Read what it did

Each change is recorded in Change Ledger with a signed evidence bundle.

Example data.

Autonomy & Rules

You decide how much it does

Set autonomy for each scope. Rules every AI must obey still apply when a scope acts on its own. Changes can be undone.

Finding · prod-eu / shop

High restarts: checkout-6d9f (7)

The change waits for someone to approve it.

256Mi → 512Mi · waiting
Pro plan

Example data.

Deterministic checks

What it watches

Autopilot watches Kubernetes and OpenShift clusters, servers and databases. Some findings have a fixed fix operation; others give advice for you to review.

High restarts: checkout-6d9f (7)advice
OOMKilledfix
Readiness probe failing: checkout-6d9f in shopadvice
HPA checkout in shop is at its ceiling of 5 replicasadvice
All 3 pods of checkout are on one nodeadvice
Cluster DNS is degradedadvice
Ingress checkout in shop routes to a Service that does not existadvice

Example data.

Guardrails

Safe by construction

The model writes the explanation, never a command or credential. For a Kubernetes fix, Zaysa reads the spec, mints a 10-minute key with edit rights in one namespace, patches, verifies, rolls back automatically if verification fails, records the change and revokes the key.

Switch AI explanations off and deterministic detectors still find problems. The detector’s own words explain them while decisions, actions, verification, rollback and audit continue.

1words, never commands
2key minted · edit · shop · 10:00
3patch
4verify
5key revoked

With AI explanations off, the detector’s own words explain the finding. The pipeline still runs.

Example data.

ON THE BUSINESS PLAN

Let it act on its own, with guardrails

Set each scope to act on its own. Autopilot still checks guardrails, your rules, policy, freeze windows and blast radius before a fixed operation runs.

Autonomous fixes

Acts within the rules you set, then verifies and records the change.

Example data.

Incident Pass

A scoped, read-only window into one cloud account that expires on its own.

Example data.

Change Check

Paste a database migration to see what it would lock, rewrite or break. SQL is read, never run.

Example data.

Replay

See what Autopilot would have done, and what happened to its changes, before it acts on its own.

Example data.

Questions

Frequently asked questions

No. The model writes the explanation shown to your team. Fixed, server-built operations carry out actions. The model never touches a credential or a command.

More in Autopilot

See what you run, and what it costs.

Free plan, no card needed.