At SIGGRAPH 2026 in Los Angeles this week, NVIDIA formally brought its Omniverse libraries under the Agent Toolkit umbrella, giving AI agents the tools to build, inspect, and prepare simulation-ready 3D environments.
New libraries for RTX sensor simulation (ovrtx), GPU-accelerated physics (ovphysx), and CAD-to-SimReady asset conversion are available on GitHub, with SideFX and PTC among the first software makers to integrate them.
"The physical AI era will be built in simulation first," said Jensen Huang, NVIDIA's founder and CEO.
The most concrete deliverable is the Blender blueprint. Demonstrated on the SIGGRAPH floor, "SimReady Blender" is a sample workflow built with Omniverse libraries and NVIDIA NemoClaw that brings agent-ready simulation capabilities directly into Blender – handling RTX sensor simulation, physics, and asset validation without pulling artists out of their existing environment.
It is now publicly available on GitHub, and for the immersive tech community, this is worth taking seriously.
Blender has one of the most vocal open-source communities in 3D development, and it is deeply resistant to vendor capture.
NVIDIA targeting it directly signals that the company knows it cannot simply build a walled garden and expect developers to move in.
Whether the integration holds up under real production conditions remains to be seen, but the intent is clear.
The SideFX partnership is the other genuinely interesting thread. SideFX is integrating ovrtx and ovphysx into Houdini's procedural content creation workflows – and Houdini is the tool the industry reaches for when environments need to be complex, controllable and built at scale. Technical artists working in robotics or industrial simulation now have a more direct path from procedural content generation to physics-tested, sensor-validated, simulation-ready output, without the friction of a separate export pipeline.
On a recent Earnings Call, Huang outlined how the company's wider vision aligns with this.
"We are accelerating all of the world's tools so that it runs on CUDA. And you could see us doing that when I work with Cadence and Synopsys and Siemens and Adobe and that is because we are trying to get all of the world's tools to run on GPUs because they already have GPUs, and it is a lot faster.
NVIDIA Is Building a Stranglehold on the Physical AI Stack
This week's announcement makes most sense when you look at what NVIDIA is assembling across the board. The hardware layer: RTX Spark and DGX Station. The simulation layer: Omniverse. The model layer: NeMo, Isaac, Cosmos. And now the workflow layer: an agent framework that ties it all together and sits inside the tools developers already use.
