The NVIDIA Agent Toolkit is Jensen Huang’s answer to the questions enterprises keep asking. How can you make your AI agents work without losing control of your data and responsibilities?
Announced at GTC 2026 in San Jose on March 16, the NVIDIA Agent Toolkit is an open-source software stack designed to help enterprises and developers build autonomous AI agents.
The barrier to widespread deployment is trust. Agents that can perform actions within enterprise systems require guardrails, but until now, guardrails have been difficult to standardize at scale.
OpenShell and safety issues
At the heart of the toolkit is NVIDIA OpenShell, an open source runtime that enforces policy-based security and privacy guardrails on autonomous agents. In NVIDIA terminology, individual agents are called “claws,” and OpenShell throttles them.
Mr. Huang explained GTC’s staking as follows: “Claude Code and OpenClaw will spark an inflection point for agents, extending AI across generations and putting inference into action. Employees will be supercharged with teams of frontier and custom-built agents to deploy and manage.”
NVIDIA is working with Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI to build OpenShell compatibility into their security tools.
research and costs
The toolkit also includes NVIDIA AI-Q, an agent search blueprint built on LangChain. It uses a hybrid architecture. The Frontier model handles the orchestration, and NVIDIA’s open Nemotron model does the research heavy lifting. According to NVIDIA, this approach can reduce query costs by more than 50% while delivering greater accuracy than the DeepResearch Bench and DeepResearch Bench II leaderboards.
This number will be important for enterprise buyers who have been plagued by consumption-based AI pricing, which seemed manageable in pilots but became a budget issue at scale.
Who is on board?
The partner list includes Adobe, Atlassian, SAP, Salesforce, ServiceNow, Siemens, Cisco, CrowdStrike, Red Hat, Box, Cadence, Cohesity, Dassault Systèmes, IQVIA, and Synopsys.
Salesforce is building a reference architecture in which employees use Slack as an orchestration layer for Agentforce agents and leverages NVIDIA infrastructure to capture data from both on-premises and cloud environments. Atlassian is integrating Agent Toolkit into its Rovo AI strategy for Jira and Confluence. ServiceNow’s “autonomous workforce of AI specialists” is built on a toolkit powered by NVIDIA AI-Q.
And Siemens launched the Fuse EDA AI agent, which uses NVIDIA Nemotron to autonomously orchestrate workflows in its electronic design automation portfolio, from design ideation to manufacturing approval. IQVIA’s adoption numbers provide a real-world data point. The company has already deployed more than 150 agents to its internal teams and client environments, including 19 of the top 20 pharmaceutical companies.
bigger changes
NVIDIA positions itself as the software infrastructure layer for enterprise agent deployments. Agent Toolkit, OpenShell, Nemotron models, and AI-Q are the components of the stack that NVIDIA wants to put under its enterprise software.
The toolkit is currently available at build.nvidia.com and supported on AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure.
See also: AI Expo 2026 Day 1: Governance and data readiness enable the agent-based enterprise
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expos in Amsterdam, California, and London. This comprehensive event is part of TechEx and co-located with other major technology events. Click here for more information.
AI News is brought to you by TechForge Media. Learn about other upcoming enterprise technology events and webinars.
