NVIDIA Donates OpenShell for AI Security

NVIDIA said it is expanding its open-source commitment at KubeCon + CloudNativeCon by donating OpenShell and contributing to Agent Substrate. The announcements suggest a larger effort to make Kubernetes the foundation for securing and running AI agents at scale.

What Happened at KubeCon + CloudNativeCon

At KubeCon + CloudNativeCon, NVIDIA said it is deepening its open-source commitment. According to NVIDIA’s blog, the company contributed NVIDIA OpenShell, an open source runtime for securing and governing autonomous agents, to the Cloud Native Computing Foundation. As a CNCF Sandbox project, OpenShell would be governed by the organizations building it into their products and strengthened by the expertise of security professionals worldwide.

NVIDIA also contributed to Agent Substrate, a Google-led open source framework for operating long-running agent workloads at scale on Kubernetes, which was accepted into the CNCF Sandbox. In addition, NVIDIA announced new components of the NVIDIA DSX AI Factory Platform: the NVIDIA Kubernetes Engine, NVIDIA Kubernetes Recipes and NVIDIA Snapshot.

The announcements arrived as Kubernetes, the open source system that manages how applications run across thousands of AI factories, has become the foundation for AI, and the demands on it are only growing, according to NVIDIA.

Why OpenShell Matters for Agent Security

OpenShell gives organizations control over what AI agents can do. According to NVIDIA, it applies deterministic controls to agent activity, so teams can enforce policies as agents use tools, access inference services and execute workloads. Its extensible architecture lets teams connect their own compute, inference providers, credential storage and monitoring tools.

This allows organizations to apply governance across different deployment environments while integrating OpenShell with their existing enterprise systems. NVIDIA framed the cloud-native community as facing a new infrastructure challenge: giving agents secure environments in which to operate.

The company said OpenShell puts security infrastructure for autonomous agents into the community’s hands to build, inspect and improve, accelerating development and hardening safeguards. According to NVIDIA, hundreds of organizations have put OpenShell to use, and the project has received tens of thousands of contributions in less than a year since launch.

The practical question is how such control scales as agents move from isolated tasks into longer workflows. NVIDIA presented OpenShell as a way to apply policy earlier and more consistently, rather than relying only on model-level behavior.

How NVIDIA Plans to Scale Agent Workloads on Kubernetes

Governing agents is only part of the challenge, NVIDIA said. Organizations also need to deploy agents at scale.

NVIDIA is contributing to Agent Substrate, a Google-originated project in the CNCF Sandbox, and ensuring it works with NVIDIA Dynamo for model inference and the KAI Scheduler for shared cluster resource allocation. This effort brings together agent execution, inference serving and resource scheduling on Kubernetes, according to NVIDIA.

Agent Substrate preserves agent environments between tasks while releasing compute when agents are idle. By suspending and resuming those environments across a shared pool of agents, it lets teams scale agent workloads without keeping dedicated compute running for every session.

NVIDIA and Google are also collaborating on an integration of OpenShell and Agent Substrate, so enterprises can secure and govern agents across clouds and on-premises systems. The two efforts are linked in the source as a way to secure agents while also making them operationally practical at scale.

What the New AI Factory Tools Mean

Three new open source projects extend the NVIDIA DSX AI Factory Platform, NVIDIA said. The NVIDIA Kubernetes Engine and NVIDIA Kubernetes Recipes are now available in public preview and provide open source building blocks for delivering managed Kubernetes services on GPU infrastructure. They automate cluster lifecycle management and provide signed, version-pinned images.

NVIDIA said ecosystem support extends their reach across AI infrastructure platforms and cloud services, including alignment by Mirantis, Netris, OpenNebula Systems, Rafay, Spectro Cloud and vCluster, together with Firmus and SK Telecom.

NVIDIA Snapshot addresses the challenge of bringing additional inference capacity online quickly. Loading models and warming up workloads can delay scaling and recovery, according to NVIDIA. The open source, modular component restores checkpointed replicas in seconds, without model reload or warmup, supporting faster startup, recovery and elasticity for inference and agentic workloads on Kubernetes.

Design partners Crusoe and Saturn Cloud evaluated Snapshot on their own workloads using NVIDIA HGX B300 systems. Crusoe, an upstream contributor to Snapshot, tested GLM-5.3 on Crusoe Cloud and reduced the time for the workload to become ready from roughly eight minutes to 31 seconds, according to NVIDIA. Saturn Cloud benchmarked Snapshot for scale-from-zero inference and reduced vLLM startup times from between two and 8.5 minutes to six and 34 seconds.

Together, these advancements mark a step forward in making Kubernetes the leading foundation for AI workloads at scale, empowering developers to build safer, more efficient AI infrastructure for the future, according to NVIDIA.

Where the Story Fits With Grafy

For readers interested in autonomous AI, the open-agent security theme has a natural conversation in Grafy. The relevant app is Grafy Operator, the browsing agent that performs web tasks, because it sits near the question of what an agent may do online.

The source material is about enterprise Kubernetes and agent governance, so the fit is more thematic than product-direct. Grafy readers can explore the topic through Grafy Operator or discuss multi-model approaches in Grafy Chat.

For developers using Grafy, the concern is practical: how to let agents act while keeping boundaries. Grafy Operator could be a place to test browsing-agent behavior, while Grafy Chat could compare how different models handle safety prompts. But Grafy’s public feature list does not include an enterprise agent-governance runtime like OpenShell.

Frequently asked questions

What did NVIDIA announce at KubeCon + CloudNativeCon?

NVIDIA announced a set of open-source contributions and new AI infrastructure pieces. According to NVIDIA, it donated OpenShell to the Cloud Native Computing Foundation and contributed to Agent Substrate. It also introduced components of the NVIDIA DSX AI Factory Platform, including the NVIDIA Kubernetes Engine, Recipes and Snapshot. The announcements concerned Kubernetes, agents and GPU infrastructure for AI factories.

How does NVIDIA Snapshot aim to improve inference scaling?

NVIDIA Snapshot is an open-source modular component designed to restore checkpointed replicas quickly. According to NVIDIA, it can restore checkpointed replicas in seconds without model reload or warmup, supporting faster startup, recovery and elasticity for inference and agentic workloads on Kubernetes. The company described evaluations with design partners, including reduced time for a workload to become ready and faster vLLM startup times.

Does OpenShell teach a smaller model how to govern agents?

No, according to the source material, OpenShell is described as an open-source runtime for securing and governing autonomous agents. It applies deterministic controls and lets organizations enforce policies, rather than teaching a smaller model governance skills. The source does not present OpenShell that way. Its focus is control, policy enforcement and integration with existing systems.

---

Source: NVIDIA Press Room — NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure