Scaling Autonomous Agents with Nutanix Agentic AI and NVIDIA Agent Toolkit

In three short years, the LLM landscape has shifted from simple dialogue to autonomous action: evolving from the chat-based interactions of the original ChatGPT, to the rise of agentic workflows powered by open-source leaders like Llama, and now to the viral, high-utility automation of OpenClaw.

OpenClaw has rapidly become a cultural and technical phenomenon in the AI community. Following its release in late 2025, the open-source framework achieved record-breaking adoption on GitHub, fueled by its transition from simple chat interfaces to autonomous agents that can execute code and manage complex, multi-day workflows. However, this viral success carries significant risk and baggage for the enterprise. OpenClaw’s local-first philosophy, while appealing to enthusiasts, has the potential to bypass essential IT governance. For an enterprise, an agent that can act independently poses existential risks: prompt infiltration can lead to unauthorized infrastructure control, and without centralized guardrails, autonomous agents can rack up massive costs through unchecked token consumption.

We’re also working together on NVIDIA NemoClaw — an open source stack that simplifies running OpenClaw always-on assistants, more safely, with a single command. As part of the NVIDIA Agent Toolkit, it installs the NVIDIA OpenShell runtime—a secure environment for running autonomous agents, and open source models like NVIDIA Nemotron. These releases shift the architectural focus from raw autonomy to managed agency. OpenShell provides out-of-process policy enforcement, so that guardrails are enforced by the developer. Key features include individual sandboxes for each agent, a policy engine and a privacy router for local inference. Moving beyond simple assistants, these agents are designed to function as digital colleagues that can reason and refine, collaborate and automate digital labor.

The Nutanix Agentic AI solution integrates with OpenShell and AI-Q and provides a turnkey operating environment to operationalize AI factories running on NVIDIA infrastructure. By integrating the Nutanix stack with specialized inferencing and gateway services, Nutanix delivers the necessary governance and security to run these autonomous agents safely at enterprise scale, moving from hardware delivery to production-ready tokens as rapidly as possible.

The true value of the Nutanix Agentic AI solution lies in its ability to bridge the gap between radical productivity and enterprise-grade safety. While self-evolving autonomous agents like OpenClaw offer a quantum leap in organizational growth, they cannot be deployed on best-effort infrastructure without observability and enterprise grade controls. Starting with a runtime is a step forward as the ecosystem continues to develop technology for deploying autonomous agents with more safety and security. By grounding these agents in the Nutanix ecosystem, we provide multi-fold advantage for the modern enterprise:

• Zero-Trust Execution at Machine Speed: The Nutanix AHV hypervisor can support VM level isolation for agents requiring a highly privileged mode of operation interacting with core system components. This offers the strongest isolation layer to manage high-risk autonomous behaviors as well as strong multi-tenant isolation for agent sandboxes.
• Sandboxing: We apply a browser Isolation model to AI. With this design, even if an agent is subverted by a malicious prompt, the blast radius can be physically contained. Your core infrastructure remains isolated from  the agent’s execution environment.
• Zero-Trust Security Model: Agents start with zero permissions. Access to resources is granted only as needed and can be strictly monitored and enforced at the infrastructure level. Agents could be running on VMs or containers accessing LLMs and tools via the MCP protocol.  Nutanix Flow Network Security allows users to visualize traffic and apply microsegmentation policies that can be applied to agents as well as tools running on VMs and containers. 
• Cost Governance and Observability: With Nutanix Enterprise AI (NAI), the architecture transitions from inference to a coordinated reasoning and execution framework. NAI provides the private infrastructure to host and serve advanced open-weights models. It functions as the primary reasoning layer, generating the high-fidelity tokens required for complex decision-making and logic. OpenShell serves as the autonomous execution layer. It consumes the outputs generated by NAI to perform programmatic actions, interface with external systems, and complete multi-step tasks within the enterprise environment. The Agentic Gateway acts as the centralized control plane for this interaction. It manages security, fine-grained access control, and the routing of token streams between the reasoning engine (NAI) and the execution agents (OpenShell).
• Sovereign Data and Agent Context: The primary bottleneck for long-running agents is managing State. For an agent to be effective, its memory must be mapped to a high-performance data stack, typically a combination of Vector Stores for semantic retrieval and Graph Databases for relational context. By grounding the agent’s memory in Nutanix Unified Storage (NUS), the State of your enterprise remains on-premises and under your control. This integration allows the infrastructure to monitor the agent’s data footprint in real-time.

By grounding these self-evolving agents in a unified stack, combining AHV and Nutanix Kubernetes Platform (NKP) based isolation, sovereign NAI inference, and the Agentic Gateway with the class-leading NUS for persistent agent memory, Nutanix enables enterprises to move beyond simply producing tokens, to operationalizing true enterprise agency. 

With Nutanix, developers and knowledge workers alike can build and and deploy agents with OpenShell and AI-Q, turning everyday professional tasks into specialized, self-improving AI assistants. This integrated AI factory provides the CISO with peace of mind that as the organization scales its digital workforce, the data remains protected, the infrastructure remains sovereign, and the business is ready for the next phase of AI growth.

©2026 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo and all Nutanix product and service names mentioned are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. Nutanix, Inc. is not affiliated with VMware by Broadcom or Broadcom. VMware and the VMware product names recited herein are registered or unregistered trademarks of Broadcom in the United States and/or other countries. All other brand names mentioned are for identification purposes only and may be the trademarks of their respective holder(s).