Skip to main content

Policy Engine

The Policy Engine lets your team apply governance rules to cloud-resource provisioning. Every time a resource is created, updated, or imported, Atmosly evaluates it against your active policies and either allows it, warns, blocks it, or routes it for approval.

Policies govern Cloud Resources — they are separate from Guardrails (which govern Kubernetes cluster actions).


The Policy Engine lives under Settings → Policy Engine. It has two tabs:

TabWhat it contains
TemplatesA catalog of built-in, ready-to-use policy templates grouped by category. This is where you create policies.
Active PoliciesThe policies currently in effect for your organization, with their scope and enforcement.
info

The Policy Engine requires the View Cloud Resource permission (the same permission that gates Cloud Resources).


How Policies Work

A policy is a governance rule with the following attributes:

AttributeDescription
CategorySecurity, Cost, Compliance, Naming, Tagging, or Approval
Rule typeThe specific check the policy performs (see below)
EnforcementBlock (Hard) — rejects a violating request; Warn (Soft) — allows it but flags a warning. Approval-category policies instead route the request for review.
Provider scopeApplies to all providers, or only AWS / only GCP
Applicable servicesThe resource types the policy applies to (empty = all types)
Cloud accountsThe accounts the policy applies to (empty = all accounts)
PriorityLower number = higher priority when multiple policies apply
ActiveWhether the policy is currently enforced

Rule Types

Governance rules:

  • Security — require encryption at rest, require encryption in transit, block public access, require deletion protection
  • Cost / sizing — limit maximum instance size, restrict storage size, restrict storage type
  • Compliance — restrict allowed regions, restrict allowed resource types, restrict engine versions, require backups
  • Naming & tagging — enforce a naming convention, require specific tags

Approval rules (route a request for human review instead of blocking outright):

  • Always require approval
  • Require approval above a cost threshold — when the estimated monthly cost exceeds a set amount
  • Require approval when tags match a condition
  • Require approval for the first resource of its kind

Creating Policies

Policies are created from templates — battle-tested guardrails you can adopt and then tune:

  1. Go to Settings → Policy Engine → Templates.
  2. Browse the templates, grouped by category. Each card shows its enforcement (Block or Warn).
  3. Select one or more templates and click Apply selected. This seeds them as active policies for your organization.

Once seeded, open a policy from Active Policies to scope and tune it.


Scoping & Tuning a Policy

From a policy's detail page you can adjust:

  • Enforcement — switch between Block (Hard) and Warn (Soft)
  • Active — turn the policy on or off
  • Priority — 1–1000 (lower = higher priority)
  • Cloud accounts — limit the policy to specific accounts (leave empty to apply org-wide)
  • Applicable services — limit the policy to specific resource types (leave empty to apply to all)
  • Rule parameters — the specific values the rule checks (e.g., the allowed regions, the required tags, the cost threshold)

You can also Clone a policy or Remove it.


Enforcement Outcomes

When a resource is provisioned, updated, or imported, each applicable policy produces a verdict:

VerdictMeaning
PassedThe resource satisfies the policy.
WarningA soft (Warn) policy was violated — the operation is allowed, but flagged.
ViolatedA hard (Block) policy was violated — the operation is rejected.

Approval-category policies route the request to a Pending Approval state, where an authorized user approves or rejects it (optionally with notes).


Where Results Appear

  • During provisioning — the Review step and violation alerts show which policies would block or warn before you submit.
  • On the resource — the Policies tab on a resource's detail page shows the latest evaluation: which policies were checked, their category, rule type, enforcement, verdict, and message, along with an evaluation history.
note

Policy evaluations are recorded at create/update/import time. The Policies tab reflects the result of the most recent operation on the resource rather than re-running checks live.