NVIDIA adds tools to help company AI agents use your data safely

NVIDIA has added new tools to its open source Agent Toolkit. They are meant to help AI agents inside companies work from the company's own data, and stay within limits the company sets. The pitch is simple: an agent is only as useful as what it knows, and only as trustworthy as the boundaries it runs within.

What changed

The main addition is NVIDIA Agent Knowledge Foundation. It helps agents find relevant company information, understand what it means for the business, and use it to make predictions and suggest next steps. NVIDIA says companies can use it without moving their data, rebuilding existing workflows or changing governance controls. Governance controls are the rules about who can see and do what.

How it works

A new library called Auto-Ontology translates raw company data into familiar business concepts, such as customers, products and orders, and maps how they relate. The foundation also includes NVIDIA Nemotron Retriever and Kumo open models, which help agents find the right information and make predictions.

For safety, OpenShell gives companies control over what agents can access and do. NVIDIA also offers NemoClaw blueprints, which are reference designs for building secure, specialized agents. The newest one is a step-by-step starting point for making agents better at their jobs over time.

Who is using it

• NVIDIA's own supply chain team uses it through Palantir to deploy specialized agents and predict outcomes.
• Databricks pairs NVIDIA Kumo Anomaly with its Genie Ontology for context-aware agents.
• Snowflake lets customers use Kumo models in its Snowflake Notebook environments.
• AWS, Dell, IBM, Oracle, Perplexity and others are also building on it.

Search results give a sense of the gains. Better search means an agent pulls the right records before it answers. SAP's fine-tuned reranker, a tool that puts the best search results first, ran nearly 3x faster. ServiceNow cut latency by nearly a quarter while improving reranking quality. Amazon OpenSearch Service improved relevance with fewer search steps and virtually no added latency.

What is not known yet

The material does not give pricing, a release schedule or how much setup work smaller businesses should expect. The speed figures come from NVIDIA's partners as reported by NVIDIA, not from independent tests. Most examples are large enterprises, so it is unclear how soon this reaches small companies.

Why it matters

The direction is clear. Business AI is moving toward agents that use your own records, such as customers and orders, inside limits you set. For any business weighing an AI assistant, the questions to ask a vendor are what data it can reach, and what it is allowed to do with it.