# Policies

> The five kinds of policy on Governance → Policies (alerting, tagging, custom recommendations, AI routing, and recommendation approvals), what each one does, and who can manage it.

Source: https://zop.dev/docs/zopnight/policies/overview

---

A **policy** is a rule you set up once, and ZopNight follows it for you from then on. Policies decide who hears about new recommendations, what tag each resource should carry, what counts as waste in your estate, which AI model a request goes to, and who must approve a change before it runs.

![Governance → Policies with five cards: Alerting, Tagging, Recommendation, AI Routing, and Recommendation Approvals, each with a short description and a Manage link](https://storage.googleapis.com/zopdev-blog-resources/1/files/originals/20261020/b594cb17-adb4-48bb-bcd7-79213c2a1a1b-policiesoverview.png)

*Governance → Policies: one card per kind of policy.*

Open **Governance → Policies** to see one card per kind. Each card shows how many policies you have (the AI Routing card shows **Routing posture & rules** instead). Click a card to manage that kind.

## The five kinds

**Alerting**

  Tell your team about new recommendations in Slack, Microsoft Teams, Google Chat, or a webhook, or open a Jira ticket for them.

**Tagging**

  Give every resource a consistent tag for cost reporting, and find resources whose tags break your rules.

**Recommendation (custom)**

  Write your own recommendation: choose resources and metric thresholds, and ZopNight raises it when they're met.

**AI Routing**

  See how the AI Gateway chooses between a cheaper and a stronger model for each request. The settings on this page don't change live traffic yet.

**Recommendation Approvals**

  Require approval before a recommendation is fixed or marked resolved, and choose who approves.

| Kind | What it does | Changes your cloud? |
|---|---|---|
| [Alerting](https://zop.dev/docs/zopnight/policies/alerting) | Sends a message or opens a Jira ticket when a new recommendation matches | No |
| [Tagging](https://zop.dev/docs/zopnight/policies/tagging) | Works out a tag value for each resource and checks existing tags | Only if you use **Apply to cloud** on the Smart Tags page |
| [Recommendation (custom)](https://zop.dev/docs/zopnight/policies/custom-recommendations) | Raises your own recommendations from metrics | No |
| [AI Routing](https://zop.dev/docs/zopnight/policies/ai-routing) | Explains how AI Gateway requests are routed; the page's own settings don't change live traffic yet | No |
| [Recommendation Approvals](https://zop.dev/docs/zopnight/policies/approvals) | Holds a fix until the right people approve it | No |

Policies for Terraform and OpenTofu code are managed separately, under **Governance → IaC Validations**. See [IaC governance](https://zop.dev/docs/zopnight/concepts/iac-governance).

## Who can manage policies

| Kind | Who can change it |
|---|---|
| Alerting, Tagging, Recommendation (custom) | The built-in **Editor** and **Admin** roles, or a custom role with the policy create, update, and delete permissions |
| AI Routing | The built-in **Admin** role, or a custom role that can update AI models |
| Recommendation Approvals | Only people with the built-in **Admin** role. A custom role can't be given this. |

Everyone with a built-in role, including **Viewer**, can open **Governance → Policies** and look. If you see "You don't have permission to view policies", ask your admin. See [Scopes](https://zop.dev/docs/zopnight/concepts/scopes) for roles.

## Good to know

- **Each policy has an on/off switch** (except AI Routing). Turning a policy off stops it without losing its settings.
- **Alerting, Tagging, and custom recommendation policies are checked before you save.** Click **Validate** first; ZopNight catches a policy that can't work before it goes live.
- **You can also manage policies from scripts or an AI assistant.** See each policy page for its [MCP server](https://zop.dev/docs/zopnight/integrations/mcp) tools.