Skip to main content
Ankra AI → Autonomy and approvals (/next/organisation/ai-autonomy) is where you decide what Ankra AI may do in your organisation without asking a person first, who approves the rest, and how to stop it. Everything on the page applies to the whole organisation.
On a new organisation, auto-remediation is off, and the default agents run only when someone starts them. Nothing changes a cluster on its own until someone turns it on here.

Before you start

  • A role that manages Ankra AI (the ai.manage permission, which owners and admins have) to stop Ankra AI or change the policy. Without it the page reads Your role can look at these settings but not change them.
  • A role that manages the organisation (organisation.manage) to change Finding problems on its own.
  • A Slack webhook under Integrations & alerts → Notification destinations if fixes are to be approved in Slack. See Notification routing.

Stop Ankra AI

The page opens with two stops, from the gentler to the harder. Both apply to everyone in the organisation straight away.
  1. Under Stop Ankra AI, press Pause autonomous actions… or Stop all AI….
  2. Enter a Reason. Everyone in the organisation can read the reason for a pause; only people who manage Ankra AI can read the reason for a stop.
  3. Press Pause autonomous actions or Stop all AI now.
You should see the pause card read Paused, or, after a stop, the stop card read Stopped and a banner across the page: All Ankra AI is stopped.

Fixing alerts automatically

When an alert fires, Ankra AI can find the cause and apply the smallest safe fix. The Auto-remediation switch turns this on for every alert in the organisation that Ankra AI is sent. Turning it on asks you to confirm with Turn auto-remediation on.
A rule set to Run on its own runs that fix with no approval at all, whatever the level. Set it only for a fix you would approve every time.
Press Save changes. You should see Saved. The next alert that fires follows this policy. Fixes waiting for approval are also listed in Needs you. ankra ai remediation policy prints the saved policy, and get_remediation_policy reads it over MCP. Both are read-only: the policy changes only in the portal. An organisation that never saved a policy reads Not set up yet. These are the platform’s starting values. in the portal. To choose which alerts Ankra AI investigates and what it does for each, set up a watch. See Watch for a problem.

Tickets and the board

Where ticket threads are posted and whether plans need an independent review are set on the ticket desk: Open the ticket desk under Tickets and the board. See Set up the ticket desk. The portal cannot change which agents work board tickets, or the identity they act as. Use designate_board_worker over MCP and ankra ai board-identity in the CLI. See Agents that work tickets.

Finding problems on its own

This section decides whether Ankra AI checks your clusters’ health by itself. What it finds is what auto-remediation and the ticket desk act on. You should see Saved. The next health check uses these settings. Who is told about what Ankra AI finds is set in Notification routes (Open Notification routes). See AI Insights.

Next steps