The new stack aims to enhance privacy and security controls, making self-evolving autonomous AI agents — known as claws — more reliable, scalable and accessible, the company said in a statement, News.Az reports.
According to Nvidia, NemoClaw integrates privacy and security mechanisms designed to support the safe development of autonomous AI agents while maintaining user control over data and operations.
RECOMMENDED STORIES
Nvidia founder and CEO Jensen Huang said OpenClaw has made advanced AI capabilities widely accessible and quickly became one of the fastest-growing open-source projects.
He described OpenClaw as a platform that could play a role similar to traditional operating systems, calling it a major step toward a new era in software development.
OpenClaw creator Peter Steinberger noted that the platform is designed to bring people closer to artificial intelligence and allow individuals to build their own AI agents. He added that collaboration with Nvidia and other partners is helping create both the tools and safeguards needed to develop powerful and secure AI assistants.
The NemoClaw stack uses Nvidia Agent Toolkit software to optimize OpenClaw through a single-command setup. It installs the OpenShell runtime, which provides access to open models and operates within an isolated sandbox environment designed to protect user data and maintain security.
This setup creates an infrastructure layer that allows AI agents to access the resources needed to perform tasks while enforcing policy-based safeguards for network access, privacy and system security.
NemoClaw also supports various coding agents and can work with open models, including Nvidia Nemotron, running locally on a user’s dedicated system. Through a privacy routing mechanism, agents can also connect to advanced cloud-based models.
By combining local and cloud-based models, the system enables AI agents to gradually develop new capabilities and perform tasks while operating within clearly defined privacy and security frameworks.
