NVIDIA Launches Open-Source Agent Toolkit With Seventeen Enterprise Partners at GTC Twenty Twenty-Six

NVIDIA Bets Big on Enterprise AI Agents

NVIDIA has launched its Agent Toolkit at GTC twenty twenty-six, positioning itself at the centre of the rapidly growing enterprise AI agent market. The open-source platform is designed to help businesses build, deploy, and secure autonomous AI agents that can independently complete complex tasks.

What the Toolkit Includes

The Agent Toolkit bundles three core components. OpenShell is an open-source runtime that enforces policy-based security and privacy guardrails, sitting between agents and infrastructure to control what agents can see, do, and access. Nemotron provides open models for research and reasoning tasks. The AI-Q Blueprint, built with LangChain, is an agentic search system that tops the DeepResearch Bench accuracy leaderboards whilst using a hybrid approach of frontier and open models that NVIDIA says can cut query costs by more than fifty percent.

Seventeen Partners Sign On

The announcement came with a wave of enterprise partnerships. Adobe plans to use the toolkit for hybrid creativity and marketing agents. Salesforce is pairing Nemotron models with its Agentforce platform for service, sales, and marketing, with Slack as the conversational interface. SAP is enabling custom agents through Joule Studio on its Business Technology Platform. Cisco will provide AI security through its AI Defense product.

ServiceNow, Atlassian, CrowdStrike, Cadence, Synopsys, Siemens, Dassault Systemes, Palantir, Red Hat, Box, Cohesity, Amdocs, and IQVIA round out the partner list.

Security First Approach

NVIDIA is collaborating with Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI to ensure OpenShell works with existing cybersecurity tools. The runtime uses process-level isolation and a privacy router to handle access control and data exposure.

Availability

Developers can access the toolkit today on build.nvidia.com, run it through major cloud partners including AWS, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure, or download it from GitHub for local use on GeForce RTX PCs and DGX systems. NVIDIA also announced NemoClaw, a companion stack that simplifies running OpenClaw-based agents with a single command.