NVIDIA is expanding its Agent Toolkit with Omniverse libraries that enable software developers to build physical AI applications and prepare 3D content for simulation using AI agents.
Announced at SIGGRAPH, the update integrates NVIDIA Omniverse libraries into the Agent Toolkit, allowing developers to automate tasks such as sensor simulation, physics modeling, and simulation-ready asset preparation. The company said the new capabilities are designed to help organizations accelerate the development of robots, autonomous systems, industrial AI and digital twins.
The Omniverse libraries are now available as open-source software on GitHub.
NVIDIA Brings Physical AI Capabilities to AI Agents
NVIDIA Agent Toolkit helps developers create AI agents that interact with software tools, workflows, and enterprise data. By adding Omniverse libraries, those agents can now work directly with 3D environments used to build and test physical AI systems.
The new libraries allow AI agents to inspect 3D scenes, identify issues, validate assets, and prepare simulation-ready environments before physical systems are deployed.
According to NVIDIA, preparing assets for robotics and industrial simulation requires more than realistic graphics. Models must include accurate materials, physical properties, sensors, object labels, and standardized structures before they can be used for AI training and testing.
“The physical AI era will be built in simulation first,” said Jensen Huang, founder and CEO of NVIDIA. “NVIDIA Agent Toolkit with Omniverse libraries brings AI agents into the 3D tools developers already use, helping build the simulation-ready worlds where robots, factories and autonomous systems are trained and tested long before they reach the real world.”
New Omniverse Libraries Target Simulation Workflows
The release includes several new Omniverse libraries designed for physical AI development.
The ovrtx library enables RTX-based sensor simulation, allowing developers to generate virtual camera, lidar and radar data for testing autonomous systems.
The ovphysx library adds GPU-accelerated physics simulation, enabling realistic interactions between objects through collision detection, friction, mass and motion.
NVIDIA also introduced CAD-to-SimReady tools that convert engineering CAD models into OpenUSD-based assets optimized for simulation environments.
Together, the libraries allow developers to automate many of the manual tasks required to prepare digital assets for robotics, manufacturing, and industrial AI applications.
Software Vendors Adopt Omniverse Libraries
Several software vendors are integrating the new libraries into their existing development platforms.
SideFX is incorporating Omniverse capabilities into Houdini workflows to help technical artists generate, validate, and prepare procedural 3D content for simulation.
PTC is integrating the libraries into its cloud-native Onshape CAD platform to connect engineering design, collaboration, and simulation workflows.
NVIDIA also announced a new Blender integration blueprint that demonstrates how developers can embed Omniverse libraries into existing 3D applications while allowing AI agents to assist with simulation preparation.
The blueprint is available through GitHub.
Local and Cloud AI Deployment
NVIDIA said applications built with the expanded Agent Toolkit can run across both local workstations and enterprise AI infrastructure.
The company highlighted support for systems ranging from RTX Spark workstations to NVIDIA DGX Station platforms, allowing developers to build and test physical AI workflows on hardware that matches their deployment requirements.
Several startups, including ForgeCAD, Lightwheel, Moonlake AI and Palatial, are also adopting Omniverse libraries to automate the creation of simulation-ready assets using AI-assisted workflows.
The Bottom Line
Simulation has become a critical part of developing robots, autonomous vehicles, industrial automation systems, and digital twins. NVIDIA’s expansion of Agent Toolkit extends AI agents beyond traditional software automation into physical AI workflows, giving developers tools to prepare, validate, and test complex 3D environments before deploying systems in the real world. As enterprises invest more heavily in robotics and industrial AI, simulation-ready infrastructure is becoming an increasingly important part of the AI development lifecycle.