NVIDIA opens early access to Factory Operations Blueprint for industrial agents

NVIDIA opened gated early access on October 21 for its Factory Operations Blueprint, a development harness for industrial agents that monitor and coordinate factory operations. Also called FOX, the system is designed around an always-on factory-manager agent that can direct work across machines and specialist agents for visual inspection, root-cause analysis, report generation and fleet management.

This is not a general release. Applicants need an NVIDIA Developer membership, an organizational email address and NVIDIA approval before receiving the software and documentation. The public enrollment page does not list a version, license, system requirements or general-availability date.

NVIDIA said agents built with FOX can connect its Cosmos and Nemotron models to factory knowledge and systems. Those systems include computer-aided design, product-lifecycle-management and simulation software, along with information-technology and operational-technology systems. According to NVIDIA, these connections let agents monitor operations in real time, match visual findings with machine logs during investigations and coordinate a response. The company also said Cosmos can adapt from examples and documentation without fine-tuning.

The blueprint builds on a reference design NVIDIA introduced in May. That design connected industrial data sources, machines, applications and robot fleets. It also included skills for model-training lifecycle tasks and agents for visual inspection, process compliance and material transport.

European factory projects offer context for these workflows, but they do not establish uniform production adoption. Siemens said an AI agent at its Erlangen electronics factory supports production planning. The agent generates scenarios for new orders and shows their effects against defined performance indicators before proposals are checked in a digital twin. Siemens estimated that its end-to-end agent-based planning approach could enable savings of up to 20%; it did not present that figure as a realized, independently verified saving.

In a separate printed-circuit-board inspection project, Siemens and NVIDIA are demonstrating an AI-based system at Erlangen. Siemens set targets of up to 85% less manual engineering effort, 40% fewer false calls and an improvement of more than 5% in first-pass yield. Those are project targets, not reported results.