Summary
NVIDIA says it is contributing its OpenShell agent-security runtime to the CNCF and adding Kubernetes tools for GPU cluster management and AI inference scaling.
NVIDIA said on Nov. 10, 2026, that it is contributing its OpenShell runtime for securing autonomous agents to the Cloud Native Computing Foundation (CNCF). The company also announced a contribution to Agent Substrate, a Google-led framework for running long-lived agent workloads on Kubernetes, and three new components for its DSX AI Factory Platform.
Contents
- OpenShell adds controls for agent activity
- Agent Substrate shares compute between tasks
- New tools target AI-factory operations
OpenShell adds controls for agent activity
OpenShell is an open-source runtime that lets organisations set controls on what AI agents can do. NVIDIA says it applies deterministic controls as agents use tools, access inference services and execute workloads. Its architecture can connect to an organisation’s chosen compute, inference providers, credential storage and monitoring tools, allowing the same governance approach to be integrated with existing systems and different deployment environments.
NVIDIA is placing OpenShell in the CNCF Sandbox, where the project is intended to be governed by organisations building it into their products and to draw on security expertise across the community. The CNCF is a nonprofit hosted by the Linux Foundation.
Agent Substrate shares compute between tasks
Agent Substrate is a framework for operating long-running agent workloads at scale on Kubernetes. NVIDIA is contributing to the project and says it is making the framework work with NVIDIA Dynamo for model inference and KAI Scheduler for shared cluster resource allocation.
The framework preserves agent environments between tasks while releasing compute when agents are idle. It can suspend and resume those environments across a shared pool, a design intended to let teams scale agent workloads without keeping dedicated compute running for every session. NVIDIA and Google are also collaborating on integrating OpenShell with Agent Substrate to govern agents across cloud and on-premises systems.
New tools target AI-factory operations
NVIDIA Kubernetes Engine (NVKE) and NVIDIA Kubernetes Recipes (NVKR) are in public preview. The open-source components provide building blocks for managed Kubernetes services on GPU infrastructure, including cluster lifecycle automation and signed, version-pinned images. NVIDIA says several infrastructure and cloud-service companies have aligned with it on integration plans for the tools.
The third component, NVIDIA Snapshot, is an open-source modular tool for restoring checkpointed inference replicas. NVIDIA says it can bring replicas back in seconds without reloading a model or warming up the workload, supporting startup, recovery and scaling for inference and agent workloads on Kubernetes.
NVIDIA reports that design partners evaluated Snapshot on their own workloads using NVIDIA HGX B300 systems. Crusoe tested GLM-5.3 on Crusoe Cloud and reported reducing the time for the workload to become ready from roughly eight minutes to 31 seconds. Saturn Cloud benchmarked scale-from-zero inference and reported vLLM startup times falling from a range of two to 8.5 minutes to six to 34 seconds. These figures describe partner evaluations on their own systems and workloads.