Nutanix has announced it will be launching the Agent Gateway as part of the Nutanix Enterprise AI 2.7 suite to provide organizations with a centralized front door for managing the way AI agents interact with LLMs and enterprise tools.
Agent Gateway is meant to secure the interactions between agents, models, and business applications while keeping governance consistent regardless of whether an organization is working with frontier models hosted in the public cloud or self-hosted private models. That flexibility matters as more companies move past AI pilots into production-scale agentic deployments, where autonomous agents are increasingly touching sensitive data and business applications to automate workflows, raising new questions around governance, security, and the cost of running these systems at scale.
As such, the feature will centralize token observability across model providers, so IT and platform teams can monitor usage, allocate spending, and get a clearer picture of where their AI budget is going. That visibility also makes it easier to spot workloads that could be shifted to self-hosted models, cutting reliance on external services where it makes sense.
From a technical standpoint, it sits as a control layer between requestors, meaning AI users and agents, and the models and Model Context Protocol servers they’re calling. Next, it applies access control policies and tool-level filtering so agents can reach enterprise resources without operating outside a governed environment. That includes granular access controls for MCP servers, centralized observability into token usage and LLM activity, detailed audit logs for every MCP request, a unified API that lets developers mix external provider models with self-hosted ones as needed, and token-based rate limiting that gives real-time visibility into usage across agents and teams.
Through the integration of Agent Gateway, Nutanix says the goal is consistent governance across agentic AI deployments, no matter where the underlying models live, with IT teams managing access policies, governance controls, and token consumption from one unified vantage point.
