NVIDIA announced NemoClaw, an open source software stack designed for the OpenClaw platform, during the GTC conference at San Jose, at claws. Jensen Huang, founder and CEO of NVIDIA, highlighted that OpenClaw represents the operating system for personal AI, comparing it to the historical impact of systems like Mac and Windows on personal computers. The initiative aims to make these agents more reliable, scalable and globally accessible, especially in environments that demand data protection.
The announcement came in the context of the rapid growth of OpenClaw, described as the most recently adopted open source project in history. NVIDIA collaborated with Peter Steinberger, creator of OpenClaw, to integrate infrastructure layers that address limitations related to security and privacy. NemoClaw uses NVIDIA Agent Toolkit to optimize the platform, installing OpenShell as an isolated sandbox environment that enforces policy-based guardrails.
The solution supports hybrid model execution, allowing agents to use open models like Nemotron running locally on dedicated user systems. A privacy router directs queries to cloud frontier models when necessary, maintaining tight controls over data and the network. Essa approach combines on-premises and cloud capabilities so agents develop new skills and perform complex tasks securely.
NemoClaw operates on several dedicated computing platforms. Usuários can run always-on agents on PCs and laptops with GeForce RTX, workstations with RTX PRO, as well as supercomputers such as DGX Station and DGX Spark. Dedicated configuration ensures continuous performance for building software, tools, and completing unattended tasks.
Simplified installation and model integration
NemoClaw installation occurs with a single command, facilitating adoption by developers and companies. The process incorporates OpenShell, which provides open templates and an isolated sandbox for secure execution.
This layer adds essential protections against data leaks and unauthorized access. Agentes gain productive access to necessary resources, but under previously defined policies.
https://twitter.com/NVIDIAAIDev/status/2033642064009957725?ref_src=twsrc%5EtfwSecurity and privacy as priorities
OpenShell enforces network, privacy, and security guardrails based on organizational policies. Isso addresses common concerns in autonomous agent deployments.
NVIDIA emphasizes that NemoClaw makes claws more scalable in enterprise scenarios. Empresas can define clear limits for agents’ actions.
The privacy router manages transitions between on-premises and cloud execution. Sensitive Dados remain protected when using advanced remote models.
Compatibility with NVIDIA hardware
Platforms like GeForce RTX support local execution of agents on personal PCs. Estações of RTX PRO work meets more intense professional demands.
DGX Station and DGX Spark supercomputers provide capabilities for highly complex tasks. Flexibility allows users to choose hardware according to their needs.
Practical demonstration at GTC
GTC attendees were able to try out NemoClaw at the build-a-claw event, held at GTC Park between March 16th and 19th. The hours varied from 1pm to 5pm on Monday and from 8am to 5pm on the following days, all in the local time zone of San Jose.
Visitors customized and deployed proactive always-on assistants. The activity highlighted the ease of creating personalized agents with integrated security.
Impact on the agentive AI ecosystem
NemoClaw strengthens OpenClaw by adding missing infrastructure for widespread adoption. Desenvolvedores access tools to build reliable wizards.
Collaboration between NVIDIA and the creator of OpenClaw accelerates the development of robust guardrails. Isso benefits both individual users and organizations.
The solution positions NVIDIA as a key provider in the advancement of autonomous agents. The focus on privacy and security meets growing market demands.