The American technology company NVIDIA on the GTC AI conference announced NemoClaw, a new integration for the OpenClaw agent platform that simplifies the deployment of secure, autonomous AI assistants. NemoClaw allows users to install the NVIDIA Nemotron stack and the new NVIDIA OpenShell runtime with a single command, providing the essential infrastructure to function with built-in privacy and security guardrails.
NVIDIA founder and CEO Jensen Huang described OpenClaw as the operating system for personal AI, marking a new renaissance in software development.
The NemoClaw system utilizes the NVIDIA Agent Toolkit to optimize OpenClaw, creating an isolated sandbox that protects data privacy while allowing agents the access they need to be productive. By using a privacy router, these agents can seamlessly bridge local NVIDIA Nemotron models with frontier models in the cloud, enabling them to learn new skills and complete complex tasks safely. According to their announcement, this setup ensures that self-evolving agents remain trustworthy and scalable across various environments.
NemoClaw is designed to run on a wide range of dedicated hardware to support always-on AI tasks. Supported platforms include NVIDIA GeForce RTX PCs and laptops, NVIDIA RTX PRO workstations, and high-performance AI supercomputers like the NVIDIA DGX Station and DGX Spark.
OpenClaw went through many changes since its launch, where the project’s creator, Peter Steinberger, rebranded the assistant and went through a rapid series of name changes to provide a stable reset for the network.
Originally launched as Clawdbot, the project briefly became Moltbot-symbolizing a lobster “molting” its shell-after a trademark request from Anthropic regarding its Claude AI.
