Nvidia and ServiceNow join forces to launch OpenClaw-inspired AI agent built for enterprises
Nvidia and ServiceNow just announced a partnership that promises to change the way companies handle AI agents in the corporate world. The project is called Project Arc, a desktop AI agent directly inspired by the success of OpenClaw, the platform that helped popularize desktop agents around the world since its launch in November.
The idea is simple, but it tackles a massive problem: how do you let employees use powerful AI agents on their own computers without putting sensitive company data at risk? And more importantly, without having to send everything to the cloud.
Think about it: most corporate AI solutions today run in the cloud. That means sensitive information has to leave the user’s computer and travel through external servers before any processing happens. For industries like healthcare, finance, legal, and government, that’s a serious — almost insurmountable — problem. Project Arc steps right into that gap, proposing a different approach where the AI agent lives on the employee’s desktop and has direct access to everything on the machine, no internet connection required to get the job done.
What is Project Arc and how does it work
Project Arc is a ServiceNow initiative developed in partnership with Nvidia that combines ServiceNow’s automation and enterprise workflow capabilities with Nvidia’s security infrastructure and agent execution platform. The result is an AI agent that can write code, run long-duration processes in the background, and interact with company systems — all running directly on the user’s computer.
According to Jon Sigler, executive vice president and general manager of AI Platform at ServiceNow, the company set out to build its own desktop AI agent. The agent can write code and run complex background processes, but there was one critical challenge: how to keep everything secure.
That’s where Nvidia comes in with a solution called OpenShell. It’s a sandboxed platform — essentially an isolated and protected environment that lets agents like Project Arc run without the risk of executing unauthorized actions on the user’s computer. OpenShell acts as a containment layer that prevents the agent from doing things it shouldn’t, like accessing files outside its permitted scope or accidentally deleting data.
Justin Boitano, vice president of enterprise AI platforms at Nvidia, explained that OpenShell is a secure runtime capable of running any agent. That includes not just Project Arc, but also Claude Codex, OpenClaw, and virtually any family of agents available on the market. That flexibility is a huge differentiator because it means companies aren’t locked into a single AI agent vendor.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The OpenClaw inspiration and why desktop agents have exploded in popularity
The inspiration from OpenClaw is no coincidence. Since the platform launched in November, it proved there’s enormous market appetite for AI agents that actually work on the desktop, fully integrated with everyday tools. OpenClaw showed that regular users could interact with sophisticated agents without needing any advanced technical knowledge, and that was the trigger that pushed both companies to accelerate Project Arc’s development with a focus on the enterprise environment.
Desktop-based agents have a clear advantage: they can access everything on the user’s machine without going through the internet. That brings speed, privacy, and a much deeper integration with each person’s actual work environment. OpenClaw helped popularize this format and quickly gained traction globally, creating an entirely new category of AI-powered productivity tools.
But with so many people using desktop agents, a dilemma emerged for companies: let employees use this technology and reap the productivity benefits, or cut off access entirely to avoid risks? Because yes, the risks are real and they’re serious.
The risks of desktop agents without proper protection
AI agents running on the desktop are incredibly useful and efficient, but they also carry serious risks. An agent without proper safeguards could, for example, accidentally upload proprietary content to the internet. Or it could make autonomous decisions that result in large volumes of data being deleted. It sounds like a movie plot, but it’s a very real concern that comes up constantly in information security team discussions.
Boitano pointed out that security teams refer to agents without safety guardrails as a lethal triad. That’s because these agents combine three dangerous characteristics at the same time: they have access to private data, they can be exposed to untrusted content, and they can be tricked into communicating with the outside world beyond the desktop system. When all three conditions coexist without any controls, the potential for damage is enormous.
This is exactly the problem Project Arc sets out to solve, combining the security of Nvidia’s OpenShell with ServiceNow’s governance tools.
AI Control Tower and Action Fabric: the security and governance layers
Beyond the OpenShell sandbox, Project Arc features two essential tools from ServiceNow that add extra layers of control and visibility.
The first is the AI Control Tower, a monitoring platform that tracks every action executed by AI agents. The Control Tower monitors the agent’s overall behavior — which files it reads, which commands it runs, and any action that might look out of the ordinary. It’s like having a real-time dashboard that shows exactly what each agent is doing at every moment. For IT managers and security teams, that kind of visibility is pure gold. 🎯
The second tool is Action Fabric, a ServiceNow software layer that connects the AI agent to the workflows, systems, and operational procedures already in place within the organization. In practice, this means Project Arc doesn’t operate in a vacuum. It understands the organization’s processes, knows the rules of the game, and respects the procedures already defined by management. When an AI agent can natively operate within a company’s established norms, adoption becomes much safer and far less risky.
The combination of these three layers — the OpenShell sandbox, AI Control Tower monitoring, and Action Fabric integration — creates an ecosystem where AI agents can be productive without being dangerous. According to Sigler, the goal is to let users run Project Arc knowing it’s fully controlled across all their computers.
Why this partnership makes perfect sense right now
Nvidia has dominated the AI hardware market for a while now, with its GPUs powering the core of virtually every major language model and inference system in the world. But in recent months, the company has been betting heavily on the idea of local AI — models and agents running directly on devices like laptops and workstations, without relying on cloud infrastructure. Project Arc is another move in that direction, but this time with a partner that deeply understands the corporate environment.
ServiceNow, for its part, is one of the most widely used platforms by large enterprises for managing workflows, support tickets, process automation, and much more. The company had already been investing in generative AI within its platform, with features that help IT, HR, and customer service teams work more efficiently. With Project Arc, that capability takes a significant qualitative leap because the AI agents integrated with ServiceNow now gain the ability to operate locally, leveraging Nvidia’s security infrastructure to process information with both speed and protection.
The enterprise market is under increasing pressure from data privacy regulations like LGPD in Brazil and GDPR in Europe, and at the same time, demand for AI solutions that boost productivity just keeps growing. Project Arc addresses both pressures simultaneously, delivering AI power without sacrificing regulatory compliance — which is a very compelling argument for any technology leader.
The role of AI agents in this new corporate era
AI agents represent a major evolution from traditional AI assistants. While a regular chatbot answers questions, an AI agent can plan, execute, and monitor tasks autonomously, interacting with multiple systems at the same time. In the context of Project Arc, this means an agent can, for example, receive a request from an employee, search through different internal databases, process everything locally, and deliver a result without human intervention at every step of the process. This is real automation, not just an automated response.
What makes all of this even more interesting is the orchestration layer that ServiceNow brings to the project. The platform already has extensive workflow logic built up over years, with deep integrations into major enterprise systems like ERP, CRM, ITSM, and many others. When you put an AI agent operating within that structure, the automation potential multiplies exponentially because the agent doesn’t need to reinvent the wheel. It simply learns to navigate an already well-structured environment packed with relevant data.
Security is the centerpiece of this entire architecture. With processing happening inside Nvidia’s OpenShell sandbox and monitored by ServiceNow’s AI Control Tower, companies can implement much more granular access control policies, defining exactly which information each agent can access and process. This is critical for highly regulated industries, where a data breach can result in millions in fines and irreparable damage to the organization’s reputation.
The bigger picture: systems like Project Arc and OpenShell are becoming essential
As more companies test and adopt desktop-based AI agents, solutions like Project Arc and Nvidia’s OpenShell are becoming increasingly important. It’s not just about having an agent that works well — it’s about having an agent that works well securely. And that distinction is what will separate companies that can adopt AI sustainably from those that will run into serious problems along the way.
The fact that OpenShell is compatible with multiple agent families, including Claude Codex and OpenClaw, shows that Nvidia is thinking about creating a security standard for the entire ecosystem, not just for a single product. That’s a smart strategic play because it positions the company as the security infrastructure provider for the entire next generation of AI agents.
The legacy of OpenClaw and what’s coming next
You can’t talk about Project Arc without giving proper credit to OpenClaw. The platform was a pioneer in showing that desktop agents could be accessible, functional, and integrated into people’s real workflows. Since its November launch, OpenClaw has built an impressive user base and generated a massive amount of feedback that helped shape what major players like Nvidia and ServiceNow are building now. It’s that virtuous cycle of innovation: someone paves the way, proves the concept, and pushes the giants to enter the game with more resources and reach.
What we can expect in the coming months is a race among major tech companies to launch solutions similar to or complementary to Project Arc. The combination of local AI with enterprise automation is one of the hottest trends in the industry right now, and anyone without an answer for it is going to fall behind. For companies already using ServiceNow platforms or Nvidia hardware, adopting Project Arc should be a relatively natural move, leveraging existing investments to make a major leap in AI-powered productivity. 🚀
Looking at the bigger picture, this partnership sends a clear message: enterprise AI is maturing. We’ve moved past the experimentation and pilot project phase and into the era of real implementation, where security, scalability, and regulatory compliance are non-negotiable requirements. Project Arc is a sign that the major tech companies have gotten that message and are building solutions that meet the real challenges of the global enterprise market.
