Skip to main content
mcp tools

Govern

tools
19
write tier
up to tier 2

What are the Govern tools for?

The `govern` MCP tools cover budgets, teams, roles, org members, tagging policies, smart tags and audit logs. Reads need no write tier. Role and member tools such as `create_role` and `add_org_member` sit at tier 1. Organisation deletion, cloud-account deletion, credential reads and org-level role assignment have no tool at any tier.

Tools

Govern MCP tools.
Tool Access Purpose
get_budget_spendreadGet current-month spend for the entities a budget targets — use this to answer 'how much has account X / this resource / this group spent so far', and to see which targets are tracking over
get_cloud_account_permissionsreadGet the per-permission IAM probe verdicts for one cloud account — use this to explain WHY discovery or a cost/recommendation feature is missing data, rather than guessing
get_teamreadGet details of a specific team including member count and resource count
list_audit_logsreadList API audit log entries showing who did what and when
list_notification_channelsreadList notification channels (Slack, Teams, Google Chat, webhooks) configured for the organisation
list_rolesreadList the roles defined in this organisation and the RBAC policies each one carries
list_smart_tagsreadList smart tags — the tags ZopNight DERIVES for a resource from this org's tagging policies, to fill gaps in the tags the cloud account actually carries
list_tagging_policiesreadList the tagging policies that govern which tags this org's resources must carry and how each tag's value is derived
list_team_membersreadList members of a team
list_team_resourcesreadList resources assigned to a team
list_teamsreadList teams in the organisation
list_usersreadList all users in the organisation
accept_smart_tagwriteAccept smart tags on a resource — promote tag keys ZopNight DERIVED from this org's tagging policies into cost attribution
create_budgetwriteCreate a monthly spend budget for ONE resource, resource group or cloud account
delete_budgetwriteDelete a budget
revoke_smart_tagwriteRevoke smart tags on a resource — return accepted tag keys to PENDING, so they stop being used for cost attribution
update_budgetwriteChange an existing budget's monthly amount and alert threshold
create_tagging_policywriteCreate a resource TAGGING policy — a rule that derives a tag onto this org's cloud resources so their spend can be attributed
update_tagging_policywriteChange an existing resource TAGGING policy — its label, its rule, or whether it is on

Roles and members are in reach at tier 1

Role and membership tools sit in this category at tier 1, the lowest write tier: create_role, update_role, delete_role, add_org_member, update_org_member, remove_org_member, add_team_member, remove_team_member and update_team_member_role. Every call runs as the user whose token the assistant holds, and the gateway checks that user’s own permissions, so an assistant can change roles and membership only as far as its user could in the app.

Every write is opt-in per organisation. The org’s mcp_write_tier starts at none, and a global MCP_WRITE_ENABLED switch, off by default, keeps every write tool inert until it is turned on.

Four things have no tool at any tier: deleting the organisation, deleting a cloud account, reading credentials, and assigning org-level roles. Assigning a role to a team member is different: add_team_member and update_team_member_role both take a role, at tier 1.

The four access levels in play

  • Read (36 tools): among them list_users, list_teams, list_roles, get_team, list_audit_logs, list_smart_tags, list_tagging_policies, list_notification_channels, get_budget_spend, get_cloud_account_permissions. Note the last: it returns the per-action IAM probe verdicts, so an assistant can tell you why a schedule will fail before it does.
  • Tier 1, metadata only (23 tools): create_budget, update_budget, delete_budget, accept_smart_tag, revoke_smart_tag, the role and member tools above, the team tools such as create_team and assign_resource_to_team, and validators such as validate_iac_policy. These change ZopNight’s own records. None of them touches a cloud resource.
  • Tier 2, reversible (33 tools): tagging, alerting, watch and IaC policies, notification channels and subscriptions, integrations and unit metrics, for example create_tagging_policy, update_tagging_policy, create_alerting_policy and create_notification_channel.
  • Tier 3, irreversible (3 tools): delete_unit_metric, delete_all_unit_metric_values and ensure_iac_ci.

Why budgets are tier 1 and not tier 0

A budget is a ZopNight record, so deleting one destroys no infrastructure. But budgets drive alerting, and an assistant that quietly raised a threshold would suppress the warning a human was relying on. Tier 1 is the smallest gate that makes that a deliberate choice.

Smart tags are proposals, not writes

accept_smart_tag and revoke_smart_tag act on ZopNight’s derived tags, virtual attributions used for cost reporting. They are never written back to your cloud provider, so accepting one changes how spend is attributed in reports and changes nothing in AWS, Azure or GCP.

faq · mcp

Govern tools: common questions

Can an assistant grant itself more access?

Not beyond its user. Role and member tools such as create_role, add_org_member and update_team_member_role sit at tier 1, but every call runs with the live permissions of the user whose token the assistant holds, so it can change roles and membership only as far as that user could in the app. Org-level role assignment, organisation and cloud-account deletion, and credential reads have no tool at any tier.

Why is creating a budget tier 1 rather than read-only?

A budget is a ZopNight record, so deleting one destroys no infrastructure. But budgets drive alerting, and an assistant that quietly raised a threshold would suppress the warning a human was relying on. Tier 1 is the smallest gate that makes that a deliberate choice.

Does accepting a smart tag write a tag into my cloud account?

No. Smart tags are ZopNight's own derived attributions, used for cost reporting. Accepting one changes how spend is attributed in reports and changes nothing in AWS, Azure or GCP.

See it fire on your bill.

Connect an account read-only in about five minutes. Findings start with the first scan.

472 rule families across 353 resource types on 22 platforms. Every threshold, metric, and IAM action is documented on these pages before you grant anything.

472 rule families documented
353 resource types covered
read-only default access level
Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console· Multi-cloud automation· Production-ready in 30 min· SOC 2 · ISO 27001· 20–60% off the bill, first month· 4 platforms · 1 console·