ovphysx-host-runtime-boundary

You can run OVPhysX simulations from a sandboxed agent by using a secure Blender MCP bridge to trigger native host-side execution, keeping the agent logic isolated from the host filesystem.

Running host-native robotics simulations requires an established host-side Blender environment with the ovphysx_host_helper.py utility installed and accessible via the Blender MCP interface.

The Blender MCP bridge handles physical AI workflows by providing a standardized interface for preflight checks, fixture preparation, authoritative pose sampling, simulation execution, and visual receipt generation.

Yes, you can generate visual GIF receipts from Omniverse physics simulations by executing the rendering tasks natively on the host machine through the configured Blender MCP bridge interface.

No, the OVPhysX host runtime boundary does not expose the host filesystem to the sandbox, as it provides a controlled bridge that isolates agent logic while executing native host-side tasks.

To start a configured demo simulation, call the Blender execute code tool with the run_configured_demo action to trigger the native OVPhysX simulation and rendering tasks on the host machine.