Singapore – VAST Data has announced an Agent Sandbox Foundation Stack, developed with NVIDIA, to give enterprises the persistent data infrastructure needed to run thousands of AI agents and sub-agents in production.
The stack combines the VAST AI Operating System with NVIDIA NemoClaw and NVIDIA OpenShell, according to the company’s announcement in Singapore on 22 October 2026. VAST says it is aimed at enterprises and AI service providers that want to deploy, operate and scale governed agent environments without tying persistent state to ephemeral compute.
The launch comes as increasingly autonomous agents change infrastructure requirements. VAST says agents can spawn other agents, execute tasks and disappear, with fleets creating and destroying thousands of isolated workspaces an hour.
Long-running agents add further complexity, as they may work across multiple sessions, sleeping, resuming and delegating tasks to short-lived subagents. In either case, VAST argues that an agent’s progress, working data and memory must persist independently of any single sandbox or compute node, and remain available, governed and auditable wherever execution resumes.
To address this, the company pairs NVIDIA’s secure agent runtime technologies with its own data services. NVIDIA OpenShell provides isolated execution and policy enforcement around agent activity, while VAST supplies persistent state, scalable agent workspaces, integrated data services and operational records across the agent lifecycle.
By decoupling persistent agent state from individual compute nodes, the architecture allows agents to pause, resume or move across eligible infrastructure without losing progress, working data or context, VAST says. The company summarises the approach as keeping agent compute ephemeral while agent state stays durable.
“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 of 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 organisations can scale agentic systems without sacrificing persistence, governance, or control,” Horev added.
NVIDIA also commented on the collaboration. “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 of NVIDIA.
“By combining NVIDIA NemoClaw and OpenShell with the VAST AI Operating System, organisations can build agent sandbox environments that are secure and designed to scale from experimentation to production,” Briski continued.
Technically, the architecture spans three layers, which VAST says removes the need to assemble separate infrastructure for agent execution, data access, persistence and observability.
The first layer covers dynamic agent environments, where the VAST AI OS sits beneath NemoClaw and OpenShell and uses zero-copy clones to provision persistent storage for isolated OpenShell sandboxes without creating full copies of the underlying data.
The second layer handles persistent state and data services. VAST DataStore provides persistent sandbox storage and snapshots, while VAST DataBase and integrated data services give isolated agents shared access to memory, context and enterprise data, so they can work concurrently without redundant copies.
The third layer focuses on observability and auditability. The VAST Management System offers real-time visibility into data flows and infrastructure activity across agent fleets, while VAST DataBase keeps a persistent record of tool calls, prompts, tokens, costs and decisions.
According to VAST, that record gives platform teams a durable audit trail across multi-agent workflows, even after an agent has disappeared.
The new stack builds on the company’s earlier Foundation Stacks, which are designed to turn NVIDIA AI Blueprints into repeatable, production-ready implementations on the VAST AI OS. The initial releases covered enterprise research assistance using VAST DataEngine with NVIDIA Nemotron, and video intelligence using NVIDIA Cosmos reasoning models and NVIDIA Video Search and Summarisation.
Looking ahead, VAST will demonstrate the stack at NVIDIA GTC Berlin through a governed multi-agent workflow running NemoClaw and OpenShell on the VAST AI OS. The demonstration will show agents spawning dynamically, accessing persistent data services and maintaining state across changing compute environments.
VAST also plans to publish the automated deployment in a public GitHub repository. That would give organisations a reproducible route to deploying the architecture in their own environments.
