NVIDIA contributes OpenShell agent-security runtime to the CNCF
OpenShell joins the CNCF Sandbox and moves from NVIDIA's GitHub to a neutral organization. NVIDIA also announced Kubernetes tools for AI factories at KubeCon.
Key takeaways
- NVIDIA contributed OpenShell, open source software for governing autonomous AI agents, to the CNCF, which welcomed it as a Sandbox project on Nov. 10.
- The code is being transferred out of NVIDIA's own GitHub repository into a neutral organization, and the CNCF post lists vendor-neutrality commitments and says OpenShell is taking up Cedar, another CNCF project, for its policies.
- NVIDIA says it is contributing to Agent Substrate, a Google-originated CNCF Sandbox project, and working with Google on an OpenShell integration, without giving a timeline.
- NVIDIA's adoption figures, such as hundreds of organizations using OpenShell, and its Snapshot benchmark results come from NVIDIA's own announcement.
- NVIDIA also announced Kubernetes Engine and Kubernetes Recipes in public preview, plus Snapshot, for its DSX AI Factory Platform.
NVIDIA has handed OpenShell, its open source software for controlling what autonomous AI agents can do, to the Cloud Native Computing Foundation (CNCF). The announcement came on Nov. 10 at KubeCon + CloudNativeCon North America. OpenShell becomes a CNCF Sandbox project, and the code is being transferred out of NVIDIA's own GitHub repository into a neutral organization named ai-openshell, according to a CNCF blog post written by CNCF's Jonathan Bryce and NVIDIA's Ali Golshan.
What does OpenShell do?
NVIDIA says OpenShell lets teams set rules, applied the same way each time, over how agents use tools, call inference services and run jobs. The CNCF post says that enforcement sits outside the agents, within the infrastructure they run on, and that OpenShell is taking up Cedar, another CNCF project, for its policy language. NVIDIA adds that the software can be wired into the compute, model-serving, secrets and monitoring systems an organization already runs.
What changes with CNCF governance?
The CNCF post lists a public RFC process, the CNCF Code of Conduct and explicit vendor-neutrality commitments among OpenShell's governance terms. It says core contributors at Canonical, Docker, NVIDIA and Red Hat have been involved since inception, and that NVIDIA's core engineering work will continue. Chris Wright, Red Hat's chief technology officer, said in the post that Red Hat is "excited to help steward OpenShell alongside NVIDIA and the broader ecosystem."
Jim Zemlin, CEO of the Linux Foundation, which the CNCF is part of, said in NVIDIA's announcement that the contribution gives developers "a shared foundation to build on." NVIDIA says hundreds of organizations have used OpenShell and that, in under a year since launch, it has received tens of thousands of contributions. Those figures come from NVIDIA, and NextFish has not verified them.
How does it connect to Google's Agent Substrate?
NVIDIA says it is also contributing to Agent Substrate, a Google-originated framework, already accepted into the CNCF Sandbox, for running long-lived agent workloads on Kubernetes. In a separate blog post dated Nov. 6, Google said it was giving three of its open source projects, Agent Substrate, gVisor and OpenRL, to the CNCF, and that Agent Substrate had reached version 1.0 and general availability on Google Kubernetes Engine. Google says the framework snapshots idle agents to storage and restores them in under 500 milliseconds.
NVIDIA says it is making sure Agent Substrate works with its Dynamo inference software and KAI Scheduler, and that NVIDIA and Google are collaborating on an integration between OpenShell and Agent Substrate. NVIDIA's post gives no timeline for that integration, and Google's post, which does not mention OpenShell, says nothing about it.
What else did NVIDIA announce?
NVIDIA also announced new parts of its DSX AI Factory Platform. NVIDIA Kubernetes Engine and NVIDIA Kubernetes Recipes are in public preview, and NVIDIA says the pair automates the life cycle of clusters and supplies images that are signed and pinned to specific versions. A third component, NVIDIA Snapshot, brings back checkpointed inference replicas without reloading the model. Among the companies NVIDIA lists as aligned with it on integration plans are Rafay, Spectro Cloud, vCluster, Mirantis, Netris, OpenNebula Systems, Firmus and SK Telecom.
NVIDIA also reports results from two design partners. It says Crusoe cut the readiness time for a GLM-5.3 workload on NVIDIA HGX B300 systems from about eight minutes to 31 seconds. It says Saturn Cloud reduced vLLM startup from between two and 8.5 minutes to "six and 34 seconds." Crusoe is also an upstream contributor to Snapshot. These benchmark figures appear in NVIDIA's own announcement and describe the design partners' tests on their own workloads.
Questions readers are asking
What is NVIDIA OpenShell?
NVIDIA describes OpenShell as an open source runtime that secures and governs autonomous AI agents, applying controls over how they use tools, inference services and workloads.
Who governs OpenShell now?
OpenShell is a CNCF Sandbox project. The CNCF post says it is moving to a neutral GitHub organization, ai-openshell, and follows governance with vendor-neutrality commitments, while NVIDIA's core engineering work continues.
What is Agent Substrate?
Agent Substrate is a Google-originated open source framework for running long-lived agent workloads on Kubernetes. It has been accepted into the CNCF Sandbox, and Google says it reached version 1.0.
Where do the adoption and benchmark numbers come from?
The claims that hundreds of organizations use OpenShell, and the Crusoe and Saturn Cloud Snapshot results, come from NVIDIA's own announcement. NextFish has not independently verified them.
Sources
- #nvidia
- #openshell
- #cncf
- #kubernetes
- #ai-agents
- #open-source
Get the day's top AI & tech stories on WhatsApp: Follow the NextFish channel
Get stories like this every morning.
Related
Agents3 min
OpenAI launches ChatGPT Wallet so its dots agents can buy for users
The payment feature starts with US ChatGPT Pro subscribers. Users approve each purchase with a biometric check, and agents get a payment token, not the real card number.
Chips & Infra3 min
Megaspeed says Alice Huang left in 2025, hires counsel for internal probe
The Singapore-based Nvidia cloud partner says reports about its supply chain lack verifiable evidence, after a WSJ report on billions of dollars in Nvidia chip orders for China.
Chips & Infra3 min
Major Nvidia investors ask board to keep door open to Huang's return
Altimeter's Brad Gerstner and Atreides' Gavin Baker want a slow CEO search, while a petition backed by Elon Musk asks the board to reinstate Jensen Huang.
Chips & Infra4 min
Researcher says Chinese open models carry Nvidia Blackwell fingerprints
Will DePue's report says six models from four Chinese labs were started on Blackwell chips. It says it cannot show where the chips were or that any law was broken.