Agents need knowledge and boundaries
NVIDIA announced an expansion of its open source Agent Toolkit with new tools. The company says enterprise agents are only useful if they can reach the right information and run within trustworthy boundaries. The toolkit aims to meet both needs by bringing together models, a runtime, data and tools.
Agent Knowledge Foundation
The new NVIDIA Agent Knowledge Foundation turns company data into enterprise intelligence that agents can use. It works without moving data, rebuilding existing workflows or changing governance controls. It includes CUDA-X libraries (cuDF, cuGraph, cuML, cuOpt, cuVS) and the new Auto-Ontology library, which translates raw data into business concepts such as customers, products and orders.
Models and a secure runtime
The foundation includes Nemotron Retriever and the Kumo family of open models (Kumo Tabular, Kumo Relational, Kumo Time Series, Kumo Forecast). OpenShell provides the secure runtime boundary, while NemoClaw blueprints offer reference designs for secure, specialized agents. A new NemoClaw blueprint focuses on agent optimization.
Who is using it
NVIDIA is using Agent Knowledge Foundation in its own supply chain through Palantir Foundry, Ontology and AIP. Databricks pairs Kumo Anomaly with Genie Ontology, and Snowflake is enabling Kumo Engine and Kumo Relational models for customers. Nemotron Retriever models are used by Amazon OpenSearch Service, SAP, ServiceNow and Zoom. AWS, IBM, Oracle, Perplexity, Dell, NetApp and Elastic are also among those building on the foundation.
