---
title: "AI Control Point"
description: "Identity-aware governance for LLM access — enforce who can call which model, at what cost, with full audit trails."
diataxis: explanation
applies_to:
  product: "nirmata-ai-agents"
audience: ["platform-engineer"]
last_updated: 2026-04-16
url: https://docs.nirmata.io/docs/controllers/mcp-ai-gateways/
---


> **In Private Preview** — AI Control Point is available to select customers. [Contact us to learn more](https://nirmata.com/request-a-demo/).

## Overview

**Nirmata AIControl** is the enforcement layer for AI — governing how developers and agents access large language models across your organization. It sits between your developers and LLM providers, evaluating Kyverno CEL policies on every request before a token is consumed.

Where billing dashboards tell you what happened yesterday, AIControl governs what is allowed right now — who can call which model, how much they can spend, and who approved any exception.

## What It Does

| Capability | Description |
|---|---|
| **Identity-aware model access** | Enforce which models each developer, team, or AD group can call — before the request reaches the provider |
| **Session budget enforcement** | Set spending limits per developer or team; degrade to a cheaper model or escalate to a manager when a budget is exhausted |
| **Cost attribution** | Tag every LLM call to a developer, team, and work item in real time |
| **Audit trail** | Full log of every request, policy decision, exception, and approval |
| **Human-in-the-loop (HITL) workflows** | Route budget exceptions to a manager for approval via Slack or other integrations |
| **Multi-provider coverage** | Govern Claude, Gemini, and OpenAI through a single policy layer via LiteLLM |

## How It Fits In

AIControl uses the same Kyverno CEL policy language as the rest of the Nirmata platform — the policies your team already uses for Kubernetes admission control now govern your AI layer. Policies are version-controlled in Git and reviewed like any other infrastructure change.

It deploys alongside LiteLLM with zero code changes to developer workflows — one configuration change routes requests through the governance layer.

## Part of the Nirmata Control Family

AIControl is one of several Policy Control Points in the Nirmata platform, all powered by the same Kyverno policy engine:

**Kubernetes Control Point → Pipeline Control Point → Terraform Control Point → Cloud Control Point → AI Control Point**

## Get Access

AI Control Point is currently available in private preview for select customers.

[Contact Us to Learn More →](https://nirmata.com/request-a-demo/)