NVIDIA has announced the launch of its Agent Toolkit, an open-source platform designed to help enterprises build, deploy, and manage autonomous AI agents capable of completing complex tasks across business workflows. The release introduces a full stack of models, agents, and runtimes—including NVIDIA OpenShell™, a secure environment that enforces policy-based guardrails for safe agent deployment.

The announcement reflects a growing move toward AI systems that execute work across enterprise environments, rather than simply generating answers. Companies including Adobe, Salesforce, SAP, and Siemens are already integrating the toolkit to build agents that operate across functions like customer service, engineering, and data analysis. This matters for organizations deciding how to scale productivity, as AI agents begin to take on tasks that previously required human coordination across multiple tools.

In short, NVIDIA is introducing an open platform for building autonomous AI agents that can reason, act, and integrate into enterprise systems. These agents are designed to independently determine how to complete tasks while operating within defined security and governance boundaries.

An AI agent is a system that can perceive information, reason through decisions, and take actions to complete tasks with minimal human intervention.

NVIDIA’s Agent Toolkit introduces a full-stack system for building secure, autonomous enterprise AI agents that can execute tasks across workflows.

NVIDIA’s Agent Toolkit is designed as a full-stack platform for building and operating enterprise AI agents, combining models, agent frameworks, tools, and runtime infrastructure into a unified system. The platform is intended to improve agent safety, security, and efficiency while supporting the growing use of AI systems that take on a more active role in software and knowledge work. It includes multiple components:

Developers can use these components to build specialized AI agents that autonomously determine how to complete assigned tasks while interacting with enterprise systems.

According to NVIDIA CEO Jensen Huang, “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.”

The AI-Q architecture enables agents to operate across enterprise knowledge systems by allowing them to perceive, reason, and act on relevant data, automatically selecting the appropriate sources and depth of analysis to deliver precise, context-aware answers.

A built-in evaluation system provides transparency into how each AI-generated answer is produced, helping developers and enterprises understand and validate agent behavior.

The AI-Q hybrid architecture combines different types of models, using frontier models for orchestration and decision-making and NVIDIA Nemotron models for research and analysis, allowing each model to handle the tasks it performs most efficiently.

This structure allows agents to combine high-level decision-making with efficient task execution across enterprise workflows.

NVIDIA reports that this architecture can reduce query costs by more than 50% while maintaining high accuracy, and that it was used to develop the top-ranking AI agent on the DeepResearch Bench and DeepResearch Bench II leaderboards.

NVIDIA is also working with LangChain, an agent engineering company whose open-source frameworks have been downloaded more than 1 billion times, to integrate the Agent Toolkit—including AI-Q, OpenShell, and Nemotron models—into the LangChain deep agent library. The LangChain integration connects NVIDIA’s agent architecture to one of the most widely used development ecosystems, enabling developers to build and scale more advanced enterprise AI agents.

As enterprises move toward deploying AI agents that can take action across systems, one of the core challenges is ensuring those agents can operate productively without compromising security, privacy, or compliance.

OpenShell is NVIDIA’s open-source runtime designed to address this challenge by giving autonomous agents the access they need to perform tasks while enforcing policy-based controls across enterprise environments.

OpenShell provides:

NVIDIA is working with security providers including Cisco, CrowdStrike, Google, Microsoft Security, and TrendAI to integrate cybersecurity and AI security tools directly into OpenShell, enabling compatibility with existing enterprise security systems.

OpenShell enables enterprises to deploy autonomous agents across workflows while maintaining compliance, governance, and operational control.

As enterprises begin deploying AI agents that can operate across workflows, major software and infrastructure providers are integrating NVIDIA’s Agent Toolkit to expand agent capabilities within their platforms.

A broad range of companies are adopting the toolkit—including OpenShell, AI-Q, and Nemotron models—to support real-world use cases across industries:

Adoption across these platforms shows that AI agents are moving into core business systems, where they are beginning to execute real work across industries rather than remain experimental tools.

Developers can explore NVIDIA Agent Toolkit and OpenShell through build.nvidia.com, where the platform is available across a range of inference providers and NVIDIA Cloud Partners.

The toolkit runs on infrastructure from providers including:

Inference and cloud partners:
Baseten, Bitdeer AI, CoreWeave, DeepInfra, DigitalOcean, GMI Cloud, Fireworks, Lightning, Together AI, and Vultr

Developers can also use OpenShell with LangChain or download it from GitHub to run locally on NVIDIA GeForce RTX-powered PCs and laptops, NVIDIA RTX workstations, and NVIDIA DGX systems—including DGX Station and DGX Spark—from hardware partners such as Altos Computing, ASUS, Dell Technologies, GIGABYTE, HP, Lenovo, MSI, and Supermicro.

For enterprise deployment, organizations can build and run AI agents on AI factory infrastructure from cloud providers including Amazon Web Services (AWS), Google Cloud, Microsoft Azure, Microsoft Security, and Oracle Cloud Infrastructure, as well as on on-premises servers from Cisco, Dell Technologies, HPE, Lenovo, and Supermicro.

Q: What did NVIDIA announce with the Agent Toolkit?
A: NVIDIA launched an open-source platform that enables enterprises to build, deploy, and manage autonomous AI agents capable of executing complex tasks across business workflows.

Q: How do NVIDIA AI agents actually work within the Agent Toolkit?
A: The system combines components like the AI-Q blueprint for orchestration, Nemotron models for reasoning, and OpenShell for runtime control, allowing agents to perceive data, make decisions, and take action across enterprise systems.

Q: Why is NVIDIA introducing AI agents now for enterprise use?
A: Enterprises are moving beyond chat-based AI toward systems that can execute tasks across workflows, making autonomous agents a key focus for improving productivity and reducing manual coordination.

Q: What role does OpenShell play in deploying AI agents safely?
A: OpenShell provides a secure runtime environment with policy-based guardrails that control how agents access data, interact with systems, and operate within enterprise security and compliance requirements.

Q: Which companies are adopting NVIDIA Agent Toolkit and how are they using it?
A: Companies including Adobe, Salesforce, SAP, Siemens, and ServiceNow are integrating the toolkit to build AI agents for workflows such as customer service, engineering design, and enterprise automation.

Q: What are the main limitations or open questions around enterprise AI agents?
A: Key challenges include ensuring reliability, maintaining governance and security, and building trust in systems that make decisions and take action independently.

This announcement highlights a growing transition in enterprise AI from tools that assist humans to systems that can take action independently.

Key point: NVIDIA is building infrastructure for AI agents that operate as active participants in business workflows rather than passive tools.

Who should care: Enterprise leaders, developers, and IT teams should pay attention because these systems directly affect how work gets done across operations, customer service, marketing, and engineering.

Why this matters now: Organizations are increasingly looking for ways to automate complex workflows, and AI agents offer a path to scaling productivity without adding human overhead.

What decision this affects: Companies must decide whether to continue treating AI as a support tool or begin integrating agent-based systems that can execute tasks autonomously.

In short, this release reflects a practical move toward AI systems that do more than generate responses—they carry out work across enterprise environments. The organizations that learn how to safely deploy and manage these agents will likely define the next phase of enterprise software.

The next competitive advantage won’t come from having AI—it will come from how effectively those agents can act within real workflows.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.