What changed
NVIDIA has expanded its open-source NVIDIA Agent Toolkit with new tools to help enterprises customize and control AI agents using company data. The update includes the NVIDIA Agent Knowledge Foundation, which enables enterprises to turn data into company intelligence agents can use. This foundation includes NVIDIA CUDA-X libraries, the new Auto-Ontology library, and open models like Nemotron Retriever and the Kumo family. The toolkit also features OpenShell secure agent runtime and NemoClaw blueprints for building secure, specialized agents.
What this means in practice
Enterprises can now use the NVIDIA Agent Knowledge Foundation without moving data, rebuilding existing workflows, or changing governance controls. The foundation helps agents find relevant information, understand its business implications, and use it to make predictions and assist in decision-making. Companies like Palantir, Databricks, and Snowflake are already using these tools to deploy specialized agents, create actionable intelligence from proprietary data, and optimize decision-making. The new NemoClaw blueprint provides a step-by-step starting point for improving agent intelligence over time.
Limitations and availability
The new NVIDIA Agent Knowledge Foundation is already in use across enterprises. However, the supplied announcement does not specify the exact availability timeline for all tools. The toolkit includes secure runtime boundaries and reference designs for building specialized agents, but the exact hardware requirements and licensing conditions are not detailed in the snapshot. The tools are available through various platforms, including Snowflake Notebook environments and Palantir Foundry, Ontology, and AIP.
