News

NVIDIA Expands Enterprise AI Push with OpenShell and Agent Software

• By John K. Waters
• March 18, 2026

NVIDIA on Tuesday unveiled an open-source software package aimed at helping companies build and run artificial intelligence agents, expanding its push to supply the infrastructure and tools behind a new wave of enterprise automation.

The toolkit, announced at NVIDIA’s GTC 2026 conference, is designed for developers and businesses creating AI agents capable of performing tasks with limited human oversight. The software includes models, agent blueprints, and a new open-source runtime called OpenShell, which applies policy-based security, network, and privacy guardrails to make autonomous agents safer to deploy.

The company framed the launch as part of a broader shift in office software and knowledge work, as businesses experiment with AI systems that can search internal data, reason through tasks, and take actions across applications.

“Claude Code and OpenClaw have sparked the agent inflection point, extending AI beyond generation and reasoning into action,” said NVIDIA CEO Jensen Huang, in a statement. "Employees will be supercharged by teams of frontier, specialized, and custom-built agents they deploy and manage. The enterprise software industry will evolve into specialized agentic platforms, and the IT industry is on the brink of its next great expansion.”

Among the new pieces is NVIDIA's AI-Q Blueprint, built with LangChain, designed to enable developers to create agents that can search enterprise knowledge, select relevant data sources, and explain how answers were produced. NVIDIA said the system uses a hybrid architecture, relying on frontier models for orchestration and its own Nemotron open models for research, a setup it said can cut query costs by more than 50% while maintaining high accuracy.

NVIDIA also said it used AI-Q to build the top-ranking agent on the DeepResearch Bench and DeepResearch Bench II leaderboards.

The company said OpenShell is being developed to work with security tools from Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI. LangChain is also working with NVIDIA to integrate parts of the toolkit, including AI-Q, OpenShell, and Nemotron models, into its deep agent library, the companies said.

NVIDIA said a broad group of software companies is working with the toolkit, including Adobe, Atlassian, Amdocs, Box, Cadence, Cisco, Cohesity, CrowdStrike, Dassault Systèmes, IQVIA, Red Hat, SAP, Salesforce, Siemens, ServiceNow, and Synopsys. The company described those efforts as spanning uses from creative and productivity software to customer support, semiconductor design, and enterprise workflow automation.

Some partners have outlined specific plans. Adobe, for example, plans to use the toolkit for long-running creativity, productivity, and marketing agents. Salesforce is working with NVIDIA software, including Nemotron models, to let customers build and deploy agents through Agentforce, with Slack serving as a conversational interface and orchestration layer. Siemens is launching a Fuse EDA AI Agent using Nemotron for workflows in electronic design automation.

Developers can access Agent Toolkit and OpenShell through NVIDIA's build site and run the software through cloud inference providers and NVIDIA Cloud Partners, including Baseten, CoreWeave, DeepInfra, DigitalOcean, Fireworks, Together AI, and Vultr, among others. NVIDIA also said OpenShell can be downloaded from GitHub and run locally on GeForce RTX PCs, RTX workstations, and DGX systems from a range of hardware makers.

The announcement adds to NVIDIA's effort to move beyond selling chips and AI servers by offering more of the software stack needed to build advanced AI systems. It also comes as major software vendors race to define how AI agents will be used inside businesses, even as many products remain in early stages.

NVIDIA said in its release that many of the products and features it described remain in various stages of development and will be offered on a when-and-if-available basis.