NVIDIA is contributing its OpenShell software to the Cloud Native Computing Foundation to give developers shared tools for controlling autonomous artificial intelligence agents, the company said on Nov. 10.
The chipmaker announced the contribution in its blog post, “NVIDIA Expands Open Source Commitment With OpenShell Donation and New AI Infrastructure,” published that day. The company said organizations building OpenShell into their products would govern it through the foundation’s Sandbox program.
The company said OpenShell lets organizations control how AI agents, software systems that perform tasks using tools and services, operate. It can enforce rules when agents use tools, access services that run AI models or execute computing tasks.
Organizations can connect their own computing resources, model providers, credential storage and monitoring tools to OpenShell, the company said. That allows them to apply controls across different environments while connecting the software to existing business systems.
The company said hundreds of organizations have used OpenShell and the project has received tens of thousands of contributions in less than a year since launch. It did not give more precise totals.
“The cloud-native community has a new infrastructure challenge: giving agents secure environments in which to operate,” said Jim Zemlin, CEO of the Linux Foundation. The Cloud Native Computing Foundation is a nonprofit within the Linux Foundation.
NVIDIA said it is also contributing to Agent Substrate, a Google-originated project accepted into the foundation’s Sandbox program. The framework runs long-running agent tasks on Kubernetes, an open source system that manages applications across groups of computers.
The company said Agent Substrate preserves agents’ working environments between tasks while freeing computing resources when they are idle. NVIDIA and Google are also working to connect it with OpenShell so businesses can control agents across cloud services and their own systems.
The company also introduced NVIDIA Kubernetes Engine and NVIDIA Kubernetes Recipes in public preview as additions to its DSX AI Factory Platform. It said the tools automate the creation, maintenance and retirement of computing clusters and provide signed software images with fixed versions.
A third component, NVIDIA Snapshot, restores saved copies of running workloads without reloading models or repeating startup preparations, the company said. Its design partners tested the software on their workloads using NVIDIA HGX B300 systems.
Crusoe reduced the time for its GLM-5.3 workload to become ready from roughly eight minutes to 31 seconds, the company said. Saturn Cloud reduced vLLM startup times from between two and 8.5 minutes to between six and 34 seconds.
The company presented Snapshot as a way to bring additional capacity for running AI models online faster and speed recovery. The results came from the partners’ workloads, and NVIDIA supplied no broader performance figures.
