Summary
In late August 2026, Nutanix made Enterprise AI (NAI) 2.8 generally available, adding an MCP Gateway that governs how AI agents connect to enterprise apps and data. Admins can set which apps, tools, and data an agent may reach, while a separate MCP Server lets agents operate Nutanix infrastructure with scoped access. The release also expands Private Inference with LoRA fine-tuning, multi-GPU tensor parallelism, batch inference, and speculative decoding.
What changed
Nutanix released NAI 2.8 (GA) with a Model Context Protocol Gateway for governing agent access to enterprise applications and data, plus an MCP Server that lets agents work with Nutanix infrastructure under scoped permissions. Private Inference gains parameter-efficient fine-tuning via LoRA, multi-GPU inference through tensor parallelism, batch inference, and speculative decoding.
Why it matters
As enterprises run agents against production systems, an MCP-layer policy gateway becomes the control point for what agents can touch, mirroring the identity- and permission-governance push from Okta and fitting Nutanix's on-prem positioning. It lets regulated buyers adopt agents on private infrastructure without granting open access.
Evidence excerpt
NAI 2.8 adds "a generally available Model Context Protocol (MCP) Gateway designed to govern how artificial intelligence agents connect with enterprise applications and data"; a separate MCP Server "lets agents work with Nutanix infrastructure without giving them free rein." Private Inference adds LoRA fine-tuning, tensor-parallel multi-GPU inference, batch inference, and speculative decoding.