TECH NEWS | VAST, NVIDIA develop secure infrastructure for AI agents
The Agent Sandbox Foundation Stack combines the VAST AI Operating System with NVIDIA NemoClaw and NVIDIA OpenShell.
VAST Data and NVIDIA announced a foundation stack for deploying AI agents in secure, isolated environments, combining persistent data infrastructure with tools designed to govern agent execution as enterprises scale autonomous AI workloads.
The Agent Sandbox Foundation Stack combines the VAST AI Operating System with NVIDIA NemoClaw and NVIDIA OpenShell. It is designed to support thousands of AI agents and subagents in production while keeping their persistent data and working state separate from temporary computing environments.
As AI agents become capable of executing tasks, delegating work to other agents and operating across multiple sessions, enterprises face the challenge of preserving data and context even when individual agents or computing environments are shut down.
The companies said the architecture separates agent execution from persistent state, allowing agents to pause, resume or move across eligible infrastructure without losing their progress, working data or context.
NVIDIA OpenShell provides isolated execution and policy enforcement around agent activity, while the VAST AI Operating System supplies persistent storage, data services and operational records across the agent lifecycle.
“Every enterprise that puts an agent into production will eventually need to answer what it did and why it did it,” said Alon Horev, VAST Data’s chief technology officer and co-founder.
“An agent might exist for minutes, spawn thousands of other agents and then disappear, but its work has to persist and its actions have to stay understandable long afterward,” Horev said.
Kari Briski, NVIDIA’s vice president of generative AI software, said enterprises need infrastructure that supports secure execution while preserving the information and context agents require throughout their lifecycle.
Three layers for production AI
The foundation stack brings together infrastructure for agent workspaces, persistent data services and operational monitoring.
The infrastructure layer uses VAST’s zero-copy cloning technology to provision persistent storage for isolated OpenShell sandboxes without creating full copies of the underlying data. The approach is designed to accommodate workloads in which agents and subagents are created and removed dynamically.
The persistent data layer separates computing resources from agent state. VAST DataStore provides sandbox storage and snapshots, while VAST DataBase and integrated data services support shared access to agent memory, context and enterprise data.
The monitoring layer combines real-time visibility through the VAST Management System with persistent activity records maintained by VAST DataBase. According to the companies, these records can include tool calls, prompts, tokens, costs and decisions, allowing platform teams to review activity across multi-agent workflows.
The architecture is intended to reduce the need for organizations to assemble separate infrastructure for agent execution, data access, persistence and observability.
Extending VAST’s NVIDIA integrations
The announcement builds on VAST’s existing Foundation Stacks, which provide repeatable implementations for enterprise AI workloads using the VAST AI Operating System and NVIDIA technologies.
Earlier examples include a research assistant using NVIDIA Nemotron models and a video intelligence implementation using NVIDIA Cosmos reasoning models and NVIDIA Video Search and Summarization.
VAST said it would demonstrate the Agent Sandbox Foundation Stack at NVIDIA GTC Berlin through a multi-agent workflow using NemoClaw and OpenShell on the VAST AI Operating System. The demonstration is intended to show agents creating workspaces dynamically, accessing persistent data services and retaining state as computing environments change.
The company also plans to make automated deployment available through a public GitHub repository so organizations can reproduce the implementation in their own environments
Full disclosure: All news articles published on the TechSabado website are written by human journalists, unless otherwise specified. Final text editing is also performed by human editors, with artificial intelligence (AI) used only to assist with additional grammar and style guide corrections..
————————————————————————-
TEN YEARS OF TECHSABADO!
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE & SPOTIFY channels.
WATCH TECHSABADO ON OUR YOUTUBE CHANNEL:
WE ARE ALSO ON SPOTIFY
WATCH OUR OTHER YOUTUBE CHANNELS:
PLEASE LIKE our FACEBOOK PAGE and SUBSCRIBE to OUR YOUTUBE CHANNEL.
