NVIDIA Revolutionizes Enterprise AI with OpenShell and AI-Q

⚡

Key Takeaways

1NVIDIA launched its Agent Toolkit at GTC 2026, aimed at securing the deployment of autonomous AI agents in enterprises.

2OpenShell, a key component, enforces security safeguards for agents, supported by partnerships with Cisco and Microsoft.

3AI-Q, integrated into the Toolkit, promises to reduce AI research costs by over 50%, attracting companies like Salesforce and Siemens.

💡Why it matters — NVIDIA is establishing itself as a key player in enterprise AI infrastructure, facilitating the secure and cost-effective integration of autonomous agents.

NVIDIA Unveils Its Agent Toolkit for Enterprise AI

On March 16, during GTC 2026 in San Jose, NVIDIA unveiled the NVIDIA Agent Toolkit, an open-source solution designed to enable businesses to deploy autonomous AI agents. This launch aims to address companies' concerns regarding data security and system control while integrating these agents into their operations.

The timing of this announcement is crucial, as although AI agents have moved beyond the pilot stage, their large-scale adoption is hindered by a lack of trust. Companies are seeking solutions to ensure that agents can operate autonomously without compromising security.

OpenShell: Securing AI Agents

At the heart of this toolkit is NVIDIA OpenShell, a runtime environment that imposes safeguards regarding security, networking, and privacy for autonomous agents. These agents, referred to as "claws" by NVIDIA, are controlled by OpenShell to mitigate any security risks.

During the conference, Jensen Huang, CEO of NVIDIA, explained that tools like Claude Code and OpenClaw allow AI to extend beyond mere generation and reasoning, moving towards concrete actions. The goal is to create a workforce of specialized agents, each managing a specific domain, and coordinated on a large scale.

NVIDIA has partnered with companies such as Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI to integrate OpenShell into their security solutions, emphasizing its role as an essential infrastructure rather than a standalone product.

AI-Q: Efficiency and Cost Reduction

The NVIDIA AI-Q, included in the Toolkit, is an agent-based research plan developed with LangChain. It employs a hybrid architecture combining advanced models for orchestration and NVIDIA's Nemotron models for intensive research. This approach could reduce query costs by over 50%, while outperforming the DeepResearch Bench and DeepResearch Bench II benchmarks.

This cost reduction is particularly appealing to companies that have seen their budgets balloon with consumption-based AI solutions.

Adoption by Large Enterprises

Many companies have already adopted the NVIDIA Agent Toolkit. Among them are Adobe, Atlassian, SAP, Salesforce, ServiceNow, Siemens, Cisco, CrowdStrike, Red Hat, Box, Cadence, Cohesity, Dassault Systèmes, IQVIA, and Synopsys. For instance, Salesforce uses Slack to orchestrate its Agentforce agents with NVIDIA infrastructure, pulling data from both on-premises and cloud environments. Atlassian is integrating the Toolkit into its Rovo AI strategy through Jira and Confluence.

ServiceNow's "Autonomous AI Specialist Workforce" is built on the toolkit alongside NVIDIA AI-Q. Siemens has launched the AI agent Fuse EDA, utilizing NVIDIA Nemotron to manage electronic design workflows. IQVIA has already deployed over 150 agents, including in 19 of the 20 largest pharmaceutical companies.

NVIDIA, a Pillar of AI Infrastructure

With these innovations, NVIDIA positions itself not only as a hardware leader but also as an essential software infrastructure for enterprise agents. The success of this strategy will depend on how quickly companies integrate these agents into their operations. The Toolkit is already available on build.nvidia.com, with support on major cloud platforms like AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure.