NVIDIA donates OpenShell to CNCF, teams with Google on AI agents

NVIDIA used KubeCon + CloudNativeCon 2026 to hand one of its homegrown projects to the community. The chipmaker is donating OpenShell, an open-source runtime for securing and governing autonomous AI agents, to the Cloud Native Computing Foundation. OpenShell gives organizations deterministic controls over what agents do — enforcing policies as they hit inference services, run workloads and pick up tools. Its architecture plugs into existing enterprise systems, so governance travels across different deployment environments.

The project arrives with a track record. Since launch, hundreds of organizations have picked up OpenShell, and the project has drawn tens of thousands of contributions. Jim Zemlin, CEO of the Linux Foundation, framed the handover as a builder's moment: the donation gives developers a shared foundation and a chance to shape how agent security evolves.

NVIDIA isn't stopping at its own code. It's contributing to Agent Substrate, a Google-led CNCF Sandbox project built to manage long-running agent workloads at scale on Kubernetes. The framework wires into NVIDIA Dynamo for model inference and KAI Scheduler for resource allocation, and it suspends and resumes agent environments on demand to cut idle compute. NVIDIA and Google are also working together to hook OpenShell into Agent Substrate, so agent operations stay secure and governed whether they run in the cloud or on-premises.

There's hardware-platform news too. NVIDIA added three new components to its DSX AI Factory Platform, aimed at making Kubernetes easier to manage and AI workloads easier to scale. The company positions the additions as a way to make Kubernetes a sturdier base for AI factories, letting developers deploy safer, more efficient infrastructure at scale.