What’s stalling broader deployment is trust. Agents that can take action inside enterprise systems need guardrails, and until now, those have been hard to standardise at scale.
OpenShell and the safety problem
The centrepiece of the toolkit is NVIDIA OpenShell, an open-source runtime that enforces policy-based security and privacy guardrails for autonomous agents. In NVIDIA’s terminology, individual agents are called “claws,” and OpenShell is what keeps them in check.
Huang framed the stakes at GTC: “Claude Code and OpenClaw have sparked the agent inflexion point – extending AI beyond generation and reasoning into action. Employees will be supercharged by teams of frontier and custom-built agents they deploy and manage.”
NVIDIA is working with Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI to build OpenShell compatibility into their respective security tools.
Research and cost
Also inside the toolkit is NVIDIA AI-Q, an agentic search blueprint built with LangChain. It uses a hybrid architecture – frontier models handle orchestration while NVIDIA’s open Nemotron models do the research-heavy lifting. According to NVIDIA, this approach can cut query costs by more than 50% while still producing accuracy that tops the DeepResearch Bench and DeepResearch Bench II leaderboards.
That figure will matter to enterprise buyers who’ve been burned by consumption-based AI pricing that looked manageable in pilots and became a budget problem at scale.
Leave a Reply