The NVIDIA Agent Toolkit is Jensen Huang’s reply to the query enterprises hold asking: how will we put AI brokers to work with out dropping management of our knowledge and our legal responsibility?
Announced at GTC 2026 in San Jose on March 16, the NVIDIA Agent Toolkit is an open-source software program stack designed to assist enterprises and builders construct autonomous AI brokers.
What’s stalling broader deployment is belief. Brokers that may take motion inside enterprise techniques want guardrails, and till now, these have been exhausting to standardise at scale.
OpenShell and the protection drawback
The centrepiece of the toolkit is NVIDIA OpenShell, an open-source runtime that enforces policy-based safety and privateness guardrails for autonomous brokers. In NVIDIA’s terminology, particular person brokers are referred to as “claws,” and OpenShell is what retains them in test.
Huang framed the stakes at GTC: “Claude Code and OpenClaw have sparked the agent inflexion level – extending AI past technology and reasoning into motion. Staff will probably be supercharged by groups of frontier and custom-built brokers they deploy and handle.”
NVIDIA is working with Cisco, CrowdStrike, Google, Microsoft Safety, and TrendAI to construct OpenShell compatibility into their respective safety instruments.
Analysis and value
Additionally contained in the toolkit is NVIDIA AI-Q, an agentic search blueprint constructed with LangChain. It makes use of a hybrid structure – frontier fashions deal with orchestration whereas NVIDIA’s open Nemotron fashions do the research-heavy lifting. In line with NVIDIA, this method can lower question prices by greater than 50% whereas nonetheless producing accuracy that tops the DeepResearch Bench and DeepResearch Bench II leaderboards.
That determine will matter to enterprise patrons who’ve been burned by consumption-based AI pricing that seemed manageable in pilots and have become a price range drawback at scale.
Who’s on board?
The companion listing consists of Adobe, Atlassian, SAP, Salesforce, ServiceNow, Siemens, Cisco, CrowdStrike, Crimson Hat, Field, Cadence, Cohesity, Dassault Systèmes, IQVIA, and Synopsys.
Salesforce is constructing a reference structure the place staff use Slack because the orchestration layer for Agentforce brokers – pulling from knowledge in each on-premises and cloud environments – powered by NVIDIA infrastructure. Atlassian is integrating Agent Toolkit into its Rovo AI technique in Jira and Confluence. ServiceNow’s “Autonomous Workforce of AI Specialists” is constructed on the toolkit with NVIDIA AI-Q.
And Siemens launched the Fuse EDA AI Agent, which makes use of NVIDIA Nemotron to autonomously orchestrate workflows in its digital design automation portfolio, from design conception by manufacturing sign-off. IQVIA’s deployment numbers supply a real-world knowledge level: the corporate has already deployed greater than 150 brokers in inside groups and consumer environments, together with 19 of the highest 20 pharma firms.
The larger shift
What NVIDIA is positioning itself because the software program infrastructure layer for enterprise agentic deployment. The Agent Toolkit, OpenShell, Nemotron fashions, AI-Q are elements of a stack that NVIDIA desires sitting beneath enterprise software program.
The toolkit is on the market now on construct.nvidia.com, with assist in AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure.
See additionally: AI Expo 2026 Day 1: Governance and knowledge readiness allow the agentic enterprise
Need to be taught extra about AI and massive knowledge from trade leaders? Take a look atAI & Big Data Expo going down in Amsterdam, California, and London. The great occasion is a part of TechEx and is co-located with different main expertise occasions, click on here for extra data.
AI Information is powered by TechForge Media. Discover different upcoming enterprise expertise occasions and webinars here.
