NVIDIA Introduces NemoClaw Stack for OpenClaw AI Agent Platform

The NemoClaw stack utilizes the NVIDIA Agent Toolkit to optimize the OpenClaw platform. It allows users to deploy autonomous agents, or “claws,” by installing Nemotron models and the NVIDIA OpenShell runtime through a single command.

The OpenShell runtime functions as an isolated sandbox, providing an infrastructure layer for agents to operate. This environment is designed to enforce policy-based security, network controls, and privacy guardrails to manage how agents access and process data.

For model execution, NemoClaw supports both local and cloud-based configurations. Agents can utilize NVIDIA Nemotron and other open models running locally on a user’s dedicated system, or access frontier models in the cloud via an integrated privacy router.

The software is compatible with a range of NVIDIA hardware platforms. This includes consumer-grade GeForce RTX PCs and laptops, professional RTX PRO workstations, and enterprise-level DGX Station and DGX Spark AI supercomputers.

NVIDIA is demonstrating the technology at the GTC Park from March 16-19. The event allows participants to customize and deploy proactive AI assistants using the NemoClaw for OpenClaw stack on dedicated local computing hardware.