Nvidia Launches Agent Knowledge Foundation for Enterprise AI
The new toolkit allows AI agents to query corporate data while maintaining existing security access protocols.
Updated on Oct. 21, 2026 in Artificial Intelligence
Live Poll
Would you prefer your company's AI tools to work directly within your existing data storage systems?
Nvidia has announced the Agent Knowledge Foundation, an enterprise toolkit designed to provide AI agents with specific business context while respecting existing data access rules. The system integrates technologies acquired from Illumex and Kumo, including a new embedding model aimed at improving search relevance for corporate data.
Why it matters
Enterprises struggle to provide AI agents with accurate, real-time access to fragmented internal data, a bottleneck that forces companies to choose between data security and AI performance. This toolkit attempts to bridge that gap by mapping business concepts directly to raw data using automated knowledge graphs.
Nvidia reports that its Nemotron Retriever embedding model achieved a 65% increase in search relevance. The foundation layer utilizes a REST API and a Model Context Protocol server to facilitate agent queries across company tables.
The players
Nvidia
A semiconductor and software company that specializes in GPU hardware, CUDA compute platforms, and generative AI infrastructure.
SAP
A multinational software corporation focused on enterprise resource planning and business process automation.
Hugging Face
A collaborative AI platform that provides open-source tools, model repositories, and infrastructure for machine learning developers.
The details
The platform employs an Auto-Ontology library to construct knowledge graphs, which are databases that store structured information by mapping relationships between entities, from raw company tables. This process translates unstructured or semi-structured corporate data into semantic business concepts. The resulting models then allow AI agents to navigate internal silos without requiring the data to be migrated or managed outside of established security protocols.
Timeline
February 2026: Nvidia acquired the startup Illumex.
June 2026: Nvidia completed the acquisition of Kumo.
July 2026: SAP finalized the acquisition of Prior Labs.
September 2026: Nvidia released Kumo Tabular models and agreed to acquire Hugging Face.
October 21, 2026: The Agent Knowledge Foundation was announced at GTC Berlin.
The Tech Race
Nvidia's release follows a period of rapid consolidation, including its $12.9 billion agreement to buy Hugging Face and significant investments from competitors like SAP. This foundation layer aims to standardize how agents interact with tabular data, directly competing with independent middleware solutions in the enterprise AI market.
Enterprise developers gain a standardized method to connect existing business tables to AI agents via a REST API. This platform currently targets organizations needing to maintain strict data access governance while enabling LLMs to perform retrieval tasks on internal information.
The takeaway
Nvidia is moving to ensure its hardware-heavy stack remains the default choice for the growing ecosystem of enterprise AI agents. Observers should track the adoption rate of the Model Context Protocol server, which will indicate if these tools are successfully scaling across diverse corporate database environments.
Further reading
Explore deeper insights into the landscape of Artificial Intelligence.
Live Poll
Would you prefer your company's AI tools to work directly within your existing data storage systems?
