Quali has added a governance and lifecycle management layer for NVIDIA NemoClaw deployments to its Torque platform, targeting organisations running autonomous AI agents across multiple teams and infrastructure environments.
Torque is designed to manage NemoClaw environments with policy enforcement, usage visibility and automated teardown, addressing operational issues that emerge once deployments extend beyond a single team. Quali is positioning the software as a control layer for organisations using NVIDIA systems on premises, in the cloud or across both.
NemoClaw is NVIDIA's open-source reference stack for autonomous AI agents. It combines the Nemotron 3 Super model, the OpenClaw agent framework and OpenShell for sandboxed execution, and has been validated on DGX Spark systems.
Quali's addition comes as companies experiment with autonomous agents that can carry out tasks across files, application interfaces and messaging systems. For IT teams, the challenge has shifted from getting those agents to run at all to controlling how they are deployed, who can access them, how costs are tracked and when environments should be shut down.
Torque is intended to sit above those deployments and manage them as governed infrastructure targets. The platform can provision full NemoClaw environments from versioned blueprints, apply policy controls and automatically retire environments when agent tasks are complete.
Control layer
The product is also intended to support multi-tenant use, with several teams running separate agent environments on shared GPU infrastructure. In practice, that means enforcing isolation between teams, assigning GPU costs to each environment and applying policy boundaries across DGX Spark, DGX Station, on-premises clusters and cloud GPU systems.
That reflects a broader issue in enterprise AI operations. While many tools support model deployment or agent development, fewer focus on the administrative layer needed when multiple business units begin using those tools at the same time.
Quali said Torque gives data scientists, developers and operations teams self-service access to governed NemoClaw environments through a portal, with the aim of reducing reliance on manual infrastructure requests. The company argues that this can preserve the isolation and security model built into NemoClaw's sandboxed design while broadening access within an organisation.
Lior Koriat, chief executive officer of Quali, outlined the company's view of the gap it sees in the market.
"NVIDIA NemoClaw is a significant step. It makes powerful autonomous AI agents accessible and secure at the individual deployment level. Torque picks up exactly where that ends. When an organization wants to scale NemoClaw across ten teams, fifty environments, multiple DGX systems, and a hybrid cloud estate, they need governance, policy enforcement, lifecycle automation, and cost control. That is what Torque delivers. NemoClaw gets agents running. Torque keeps them governed," said Koriat.
Broader push
The launch also expands Quali's alignment with NVIDIA's software and infrastructure stack. Its platform already supports DGX Spark, DGX Station, Nemotron 3, NVIDIA GPU Operator and NIM Operator, and now extends that support to NemoClaw deployments.
This matters because enterprises adopting AI systems often piece together several layers of technology, including models, runtimes, orchestration tools and the underlying compute estate. As those layers become more complex, governance software is becoming a more visible part of the market, particularly for businesses trying to balance internal demand for AI experimentation with cost controls and security requirements.
OpenClaw, the open-source agent framework included in NemoClaw, has gained attention among developers as a self-hosted tool for autonomous task execution. Quali cited its rapid uptake in the open-source community as one reason enterprises are moving beyond limited pilots to broader internal use cases that require tighter oversight.
For Quali, the announcement places Torque in a part of the AI stack focused less on model development and more on operational control. Its argument is that as autonomous agents move into mainstream internal use, the key question for large organisations will be not only which models or frameworks they choose, but how those environments are governed once dozens of teams begin using them.
Quali said Torque can provide visibility across usage and infrastructure for NemoClaw environments while applying policies consistently across shared GPU resources.