Skip to main content

Overview

The Remediations page shows fixes that Niro’s AI investigator has suggested based on active incidents in your fleet. Every remediation requires your explicit approval before it runs — Niro never takes action on a cluster without a human in the loop.
Remediations are available on Pro and above. The AI investigator must be enabled to generate remediation suggestions.

Where remediations come from

When you run an AI investigation on an incident, the investigator may suggest concrete fixes — for example, restarting a crash-looping pod, scaling up a deployment, or rolling back to a previous image. These suggestions appear on the Remediations page as pending items. Suggested remediations are specific to the incident context. Each one shows:
  • The incident that triggered the suggestion
  • The proposed action and its target (which pod, deployment, or workload)
  • The reasoning behind the suggestion

Approving a remediation

1

Open the Remediations page

Click Remediations in the sidebar.
2

Review the suggestion

Read the proposed action and the AI’s reasoning. Check the target cluster and workload before approving.
3

Approve

Click Approve. Niro sends the action to the cluster’s agent immediately.
4

Verify the result

The remediation shows its outcome — success or failure — within a few seconds. The linked incident updates with the action taken.

Rejecting a remediation

If you don’t want to apply a suggestion, click Reject. The remediation is dismissed and does not run. Niro logs the rejection in the audit log.

Audit trail

Every remediation action — approved or rejected — is logged in the audit log with:
  • Who approved or rejected it
  • When the action ran
  • What the outcome was
This makes the remediation history fully reviewable, even for teams with multiple operators.

Available actions

The types of remediations Niro can suggest depend on what the AI investigator identifies. Common examples include: The agent capability to apply changes must be enabled on the cluster for any remediation to run.

Guardrails

Niro applies several guardrails before running any remediation:
  • Approval required — no action runs without explicit approval from a logged-in user
  • Cluster must be online — the agent must be active and heartbeating
  • Apply capability required — the cluster must have the apply capability enabled
  • Scope-limited — actions target only the specific resource identified in the investigation
Remediations act on live clusters. Review the target workload and cluster before approving, especially on production clusters.

AI Investigation

How investigations generate remediation suggestions.

Alerts & Incidents

Incidents that trigger investigation and remediation workflows.

Agent Capabilities

The apply capability required to run remediations.

Plans

Remediations are available on Pro and Enterprise.
Last modified on June 19, 2026