Governing AI
Organization administrators use Settings > Organization > AI Governance to control reusable AI instructions and organization-level AI usage.
Available controls
| Tab | Purpose |
|---|---|
| AI Skills | Manage reusable skill definitions and their supporting files |
| AI Requirements | Define reusable requirements that guide generated work |
| AI Usage | Review usage analytics; visible only to authorized administrators and owners |
| AI Settings | Configure daily credit limits and the organization-wide AI kill switch |
Governance practices
- Assign an owner and review date to every skill and requirement.
- Keep instructions specific, testable, and free of secrets or customer data.
- Review generated changes before accepting them into an application.
- Limit MCP tools to approved servers and least-privilege scopes.
- Monitor usage for unexpected growth, repeated failures, or abnormal tool calls.
- Use the kill switch when risk cannot be contained through narrower access controls.
The kill switch is an emergency control, not a substitute for identity, authorization, prompt review, data classification, or tool-level restrictions.
See the NxAgent user guide for AI-assisted development and AI integrations, tools, and authorization for runtime boundaries.