DocsBook a demo

Platform

Platform overviewArchitectureWorkflow orchestrationGitOps configurationGovernance and AAAAI and MCPKnowledge and contextRuntime and executionEvidence and monitoring

Use cases

All use casesProduction incidentRelease preparationHotfix to productionSecurity scan triage
Why NopsAIIntegrationsSecurity

Resources

All resourcesAI agent governanceMCP governanceMCP securitySelf-hosted platforms
PricingGitHub

Company

How a run worksAboutContactBook a demo

Governance levels

How NopsAI acts on an AI policy review decision, and why the default is the strict one.

ReferenceAdministratorSecurityAutomation author

Key points

  • There are two levels: advisory and strict. An unset governance_level normalizes to strict.
  • strict proceeds only on a clear allow. Violations, conflicts, uncertainty, and unsupported decisions all block.
  • advisory logs and warns; execution proceeds despite an evaluated policy concern.
  • Advisory never skips the evaluation. If no judgment can be obtained at all, because no model is available or the review call fails, the task fails closed at every level.
  • Any value other than advisory or strict is a validation error.

Field reference

advisorygovernance levelenforcementOptionalNot the default

Log and warn on an evaluated policy concern. Execution proceeds.

Example

governance_level: advisory

Rules
  • Advisory downgrades a judgment the model made. It never skips the evaluation: if no judgment can be obtained, the task still fails closed.
Evidence

pkg/models/policy_merge.go

strictgovernance levelenforcementOptionalDefault when unset

Proceed only on a clear allow. Violations, conflicts, uncertainty, and unsupported decisions all block.

Example

governance_level: strict

Rules
  • An empty value normalizes to strict.
  • Any value other than advisory or strict is a validation error.
Evidence

pkg/models/policy_merge.go

Examples

Strict pipeline with one advisory taskyaml
governance_level: strict
knowledge_context:
  - kind: policy
    ref: platform/deploy-policy
    required: true
steps:
  - name: analyze
    tasks:
      - name: draft-summary
        governance_level: advisory
        goal: Draft a human-readable summary of the proposed change.

How it works

The level can be set at pipeline, step, or task, with the most specific value winning. This lets a pipeline stay strict overall while one well-understood task runs advisory.

Governance interacts with Knowledge Context: the policy review is what produces the decision, and the governance level is what NopsAI does with it. Without blocking knowledge there is nothing to enforce.

Limits

Current behavior
  • Lowering the governance level weakens fail-closed enforcement; it does not disable the policy review itself.

Implementation evidence

  • pkg/models/policy_merge.go

    Level constants, normalization, and the strict default.