The new Agent Sandbox Foundation Stack combines NVIDIA’s agent technologies with VAST’s data infrastructure to help enterprises preserve agent memory, working data and activity records across temporary computing environments. 

[For more news, click here]

An AI agent may run for only a few minutes, delegate work to other agents and then disappear, while the information it has gathered and the progress it has made may be needed much later. As enterprises deploy more autonomous systems, they need ways to create isolated environments quickly while ensuring that agents can resume unfinished work without losing their data or context. 

VAST Data and NVIDIA announced a new Agent Sandbox Foundation Stack on October 22, 2026, designed to address this challenge. Combining the VAST AI Operating System with NVIDIA NemoClaw and NVIDIA OpenShell, the stack is intended to help enterprises and AI service providers deploy and scale governed agent environments without tying persistent state to individual compute nodes. The companies say the architecture is designed to support thousands of agents and subagents, with fleets creating and removing thousands of isolated workspaces an hour. 

The system separates agent execution from the data and context required to complete a task. NVIDIA OpenShell provides isolated execution and policy enforcement around agent activity, while VAST supplies the persistent storage, shared data services and operational records needed throughout an agent’s lifecycle. This allows agents to pause, resume or move across eligible infrastructure without losing their progress when their original computing environment is removed. 

Keeping Agent Work Intact 

VAST uses zero-copy clones to provision persistent storage for new OpenShell sandboxes without making full copies of the underlying data. VAST DataStore provides persistent sandbox storage and snapshots, while VAST DataBase and integrated data services support shared access to agent memory, context and enterprise information. This allows isolated agents to work concurrently with common datasets without creating redundant copies. 

The approach is intended for workflows in which agents operate for different lengths of time, return to unfinished tasks or delegate work to subagents. Their computing environments can be removed when no longer needed, while the data associated with their work remains available. 

"Every enterprise that puts an agent into production will eventually need to answer what it did and why it  did it," said Alon Horev, CTO and Co-Founder at VAST Data. "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. Together with NVIDIA, we're showing how the VAST AI Operating System provides the durable foundation underneath secure agent environments, so organizations can scale agentic systems without sacrificing persistence, governance or control." 

The stack also provides tools to monitor agent activity and retain records after tasks have ended. The VAST Management System offers real-time visibility into data flows and infrastructure activity across agent fleets, while VAST DataBase maintains a persistent record of tool calls, prompts, tokens, costs and decisions. These records are intended to help platform teams follow activity as it happens and review it after an agent has finished running. 

Monitoring and Deploying Autonomous Agents 

NVIDIA’s role centers on secure agent execution and policy enforcement, while VAST provides the infrastructure for persistent state and operational oversight.  

“As AI agents take on more complex work, enterprises need infrastructure that can support secure execution while preserving the data, context and state agents depend on across their lifecycle,” said Kari Briski, Vice President of Generative AI Software at NVIDIA. “By combining NVIDIA NemoClaw and OpenShell with the VAST AI Operating System, organizations can build agent sandbox environments that are secure and designed to scale from experimentation to production.” 

The announcement extends VAST’s Foundation Stack program, which is designed to turn NVIDIA AI Blueprints into repeatable, production-ready implementations on the VAST AI Operating System. Earlier stacks demonstrated enterprise research assistance using NVIDIA Nemotron and video intelligence using NVIDIA Cosmos reasoning models and NVIDIA Video Search and Summarization. The new stack focuses on the infrastructure needed to operate autonomous agents in isolated, governed environments. 

VAST plans to demonstrate the Agent Sandbox Foundation Stack at NVIDIA GTC Berlin through a governed multi-agent workflow running NVIDIA NemoClaw and OpenShell on the VAST AI Operating System. The demonstration is expected to show agents spawning dynamically, accessing persistent data services and maintaining state as computing environments change. VAST also plans to make automated deployment available through a public GitHub repository, giving organizations a reproducible path to deploy the architecture in their own environments. 

The announcement outlines how the companies intend to bring isolated execution, persistent storage and activity records together for enterprise AI deployments. The practical value will depend on how the stack performs in individual environments, but its purpose is to let agents operate in temporary sandboxes while keeping their work and activity records available after those sandboxes are gone. 

Related Articles 

NVIDIA Just Posted Its Biggest Quarter Ever, Its CFO Says the Build Has Barely Started 

Meta Is Scaling AI With AWS Graviton, Here Are the Chips Powering the Next Wave of Agentic AI 

SandboxAQ Targets the Next Agentic AI Problem: Trust, Cost and Control