# NVIDIA contributes OpenShell to the CNCF: an inspectable runtime for agents, plus a Kubernetes substrate - Category: 开源生态 - Published: 2026-11-12T02:22:02.334Z - Source date: 2026-11-10 - HTML: https://www.88z.ai/en/news/news-nvidia-contributes-openshell-to-cncf-20261110 - Markdown: https://www.88z.ai/en/news/news-nvidia-contributes-openshell-to-cncf-20261110.md - Other language: https://www.88z.ai/news/news-nvidia-contributes-openshell-to-cncf-20261110.md On 10 November 2026 at KubeCon, NVIDIA said it is contributing OpenShell, a runtime for securing and governing autonomous agents, to the Cloud Native Computing Foundation as a CNCF Sandbox project. It is also contributing to Agent Substrate, a Google-led project already accepted into the sandbox, and announced new DSX AI Factory Platform components for deploying and scaling GPU inference on Kubernetes. ## At a glance - OpenShell is being contributed to the CNCF as a Sandbox project, giving autonomous agents a runtime that can be inspected and governed. - NVIDIA is also contributing to Google-led Agent Substrate and plans to integrate the two, so agents can be constrained across environments. - Officially, OpenShell applies deterministic controls as agents use tools, reach inference services, and run workloads. - The DSX platform also adds NVIDIA Kubernetes Engine, Kubernetes Recipes, and Snapshot; Snapshot startup figures come from partner evaluations. ## A KubeCon hand-off that puts the agent runtime under community governance On 10 November 2026, NVIDIA announced at KubeCon + CloudNativeCon that it is contributing NVIDIA OpenShell to the Cloud Native Computing Foundation, the Linux Foundation nonprofit that hosts the conference. Officially, OpenShell is an open source runtime for securing and governing autonomous agents. As a CNCF Sandbox project, it will be governed by the organizations building it into their products and opened to review by a wider group of security professionals. Linux Foundation CEO Jim Zemlin is quoted saying the cloud-native community has a new infrastructure challenge: giving agents secure environments in which to operate, and that contributing OpenShell to the CNCF gives developers a shared foundation and a chance to shape how agent security evolves. That is a quoted endorsement of the open-source intent, not a promise about how the software will behave in a reader’s cluster. ## What OpenShell governs, and what Agent Substrate adds According to NVIDIA, OpenShell gives organizations control over what AI agents can do. 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 is meant to let teams connect their own compute, inference providers, credential storage, and monitoring tools, and to apply governance across deployment environments while integrating with existing enterprise systems. The post says hundreds of organizations have put OpenShell to use, and that the project has received tens of thousands of contributions in less than a year since launch. NVIDIA is also contributing to Agent Substrate, a Google-originated open source framework for operating long-running agent workloads at scale on Kubernetes, already accepted into the CNCF Sandbox. NVIDIA says it is making sure the framework works with NVIDIA Dynamo for model inference and with KAI Scheduler for shared cluster resource allocation. A key design is that Agent Substrate preserves agent environments between tasks while releasing compute when agents are idle, suspending and resuming those environments across a shared pool so teams do not have to keep 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. ## New Kubernetes pieces for AI factories, and benchmark figures that need a source label The announcement also introduces three new open source components of the NVIDIA DSX AI Factory Platform: NVIDIA Kubernetes Engine, NVIDIA Kubernetes Recipes, and NVIDIA Snapshot. Officially, they help teams deploy, manage, and scale GPU infrastructure on Kubernetes for AI factories. Snapshot is described as restoring checkpointed replicas in seconds without a model reload or warmup, to support faster startup, recovery, and elasticity for inference and agentic workloads. Partner results are listed separately and should be labeled as such. Crusoe, an upstream contributor to Snapshot, tested GLM-5.3 on Crusoe Cloud and said time-to-ready fell from roughly eight minutes to 31 seconds. Saturn Cloud, described as a token factory platform, benchmarked Snapshot for scale-from-zero inference and said vLLM startup times fell from between two and 8.5 minutes to between six and 34 seconds. Those figures come from the partners’ own workloads on NVIDIA HGX B300 systems. They are not independent tests by this site, and they should not be copied onto every cluster. ## 88Z perspective: split policy enforcement from scheduling elasticity, then verify on your own cluster If you already run agents on Kubernetes, it helps to split the problem in two. One part is whether an agent’s tool use, credential access, and inference calls are constrained by deterministic policy and an audit trail. The other is whether long-running jobs keep GPUs reserved while idle. OpenShell and Agent Substrate are aimed at those two parts; the integration is the next step NVIDIA says it is pursuing. Adoption should depend on whether they can attach to your existing policy, secret, and monitoring stack, not on the Sandbox label alone. The Snapshot speed-ups are striking, but they should be compared with your own model-loading, image-distribution, and cold-start path before you follow them. CNCF Sandbox status also means the project is still early, so governance, interfaces, and compatibility may change. A sensible start is NVIDIA’s KubeCon material and the CNCF project pages, then a test environment that checks whether policy blocks and suspend-resume behave the way you expect. ## Sources - [NVIDIA](https://blogs.nvidia.com/blog/open-source-openshell-infrastructure) (2026-11-10) This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.