Nvidia has launched open-source AI agent “ NemoClaw” at GTC 2026, demonstrating the US-based chipmaker’s shift towards agentic AI in the midst of OpenClaw hype.
Until now, the tech landscape has been gripped with the surging popularity of OpenClaw, an autonomous agent that acts as an assistant to perform multiple tasks. Even OpenClaw mania has reached China and different Chinese companies are introducing their versions of Claws.
During the keynote address, the CEO of Nvidia Jensen Huang said, “OpenClaw opened the next frontier of AI to everyone and became the fastest-growing open source project in history.”
“Mac and Windows are the operating systems for the personal computer. OpenClaw is the operating system for personal AI. This is the moment the industry has been waiting for — the beginning of a new renaissance in software,” he added.
Despite OpenClaw’s growing popularity, the open-source AI agent has been flagged due to security reasons. According to the cybersecurity experts, OpenClaw is capable of accessing private data, communicating externally and making it vulnerable to harmful content, which the researchers called it “lethal trifecta.”
Here what makes Nvidia’s newly unveiled NemoClaw more distinguished is its ability to deliver more security and privacy.
The NemoClaw platform will allow companies to deploy AI agents, helping the firms to automate complex and multistep tasks for their entire workforce.
Nvidia NemoClaw is an open-source software stack designed to enhance the OpenClaw platform with advanced privacy and security controls.
NemoClaw consists of multiple components. The first element is OpenShell that is responsible for enforcing policy-based guardrails and giving users strict control over agent behaviour and data handling.
The other component is the new open-source tool that OpenShell powers, is an existing Nvidia project called Nemotron. It comprises more than a half dozen AI models that help in different tasks such as assessing graphs and generating texts.
By using a privacy router, the agents can use frontier models running in the cloud. The integration of local and cloud models helps the agents to learn new skills and finish tasks according to defined privacy-related instructions.
NemoClaw allows users to deploy “self-evolving” agents across various environments with a single command.
The stack optimizes local compute resources to run high-performance models locally, reducing costs and increasing data privacy.
The model is currently available in early preview.