Nvidia has developers contributing to OpenClaw full-time and is building enterprise agent blueprints through its NemoClaw framework, according to Nader Khalil, Nvidia’s Director of Developer Technologies, in an interview with The New Stack published June 21.

“We have a couple of developers at the company that contribute to OpenClaw full time,” Khalil told The New Stack. The commitment follows CEO Jensen Huang’s earlier public support for the open-source agent platform.

What Nvidia Is Actually Building

Khalil described NemoClaw as Nvidia’s “blueprint” system for helping enterprises adopt agent harnesses. “When we see amazing harnesses, we try to figure out how we can help enterprises adopt them,” he said. The blueprints set up agent runtimes, configure security policies for local GPUs, and connect to Nvidia’s model infrastructure.

“There’s a blueprint for Hermes and a blueprint for OpenClaw,” Khalil told The New Stack. The blueprints are part of Nvidia’s CUDA X library strategy, which now requires every Nvidia product to ship with a corresponding skill for agent harnesses.

Nvidia is also building OpenShell, a security runtime designed to address enterprise risk concerns around autonomous agent execution.

Enterprise Adoption Is Already Happening

Khalil named CrowdStrike, Cadence, and Palantir as companies Nvidia is already working with on agent deployments. He framed enterprise specialization as inevitable: “Every industry in enterprise will be building these specialized agents, and many already have.”

His analogy: a specialized agent should work like your own microwave. “When you use a microwave that you haven’t used before, you have to press a lot of buttons or spend time figuring it out. But when it’s your microwave at home, you just go ‘Boop, boop. Done.’”

Contributing to OpenClaw’s PR Backlog

Khalil acknowledged OpenClaw’s growing pains with unresolved pull requests across its 800,000+ line codebase. “We saw Peter tweeting about some of the issues they had, and we just rolled up our sleeves and were eager to help,” he said, referring to OpenClaw creator Peter Steinberger.

The PR bottleneck, Khalil noted, is structural: “It is easier to enlist many agents to help write code and build these PRs. The bottleneck is in merging the PRs through.” He called OpenClaw “a major change for the industry,” noting the project accumulated more GitHub stars than Linux in months.

Nvidia’s Definition of an Agent

Khalil offered a framework for understanding agents that centers on the harness, not just the model. “An agent is an LLM and a harness. And if you think about that, it involves two things. It involves the loop and the LLM,” he said. “Each loop should take us closer to our goal.”

He credited ChatGPT’s innovation as happening “outside of the model,” through system prompts, memory, and multimodal inputs. The agent harness, in Khalil’s framing, is where the real product differentiation occurs.

The Infrastructure Play

Nvidia’s approach is to fit into existing developer workflows rather than replace them. “Our goal is to create the tooling that’s needed in the ecosystem,” Khalil said. “Developers in industry and enterprises have actually been adopting agents. And we have been building for this audience.”

On concerns about long-running autonomous agents, Khalil said the anxiety is “slowly petering out.” He positioned Nvidia as focused on safe delivery: “There are gonna be some people quick to adapt. And some people that aren’t. So there’s much work in helping make sure that we deliver this safely.”