Skip to main content
This page covers deployment-level automation controls. For authoring and using automation, see Agents and Workflows.

​
Runtime modes

Set AUTOMATION_RUNTIME_MODE or Helm automation.runtimeMode to:
  • off to stop new dispatch while preserving durable state,
  • shadow to create and inspect durable intent without dispatch,
  • canary to dispatch only named canary workspaces,
  • on to dispatch all eligible workspaces.
Keep the runtime off while required migrations and startup upgrades complete. Progress through shadow and canary before enabling all workspaces.

​
Canary scope

Set AUTOMATION_CANARY_WORKSPACE_IDS or automation.canaryWorkspaceIds to the exact workspaces eligible during canary rollout. Canary selection does not bypass workspace readiness, authorization, target compatibility, or integration checks.

​
Runtime and retention

Workflow execution duration and report retention are deployment policies: Workflow definitions cannot increase these limits.

​
Diagnostics

Authorized workspace readers can inspect the automation diagnostics endpoint for runtime state, pending dispatch age, run counts, scheduler lag, pending approvals, template readiness, and retained report-source counts. Monitor the control-plane automation metrics and supplied Prometheus alert rules. Alert on delayed dispatch, scheduler lag, stale approvals, review-blocked runs, trigger rejection, MCP readiness failures, cleanup retries, report failures, and elevated terminal failures. Diagnostics and metrics must contain bounded IDs, states, counts, and reasons only. Do not export prompts, chat bodies, tool arguments, webhook payloads, reports, credentials, or continuation state.