At Nvidia GTC, the recent global AI conference, Nvidia revealed its own open source stack.
Dubbed NemoClaw, Nvidia explains that the stack “simplifies running OpenClaw always-on assistants”, but what does that really mean? How do NemoClaw and OpenClaw really work together?
To help you understand, we explain everything we know about Nvidia’s NemoClaw including what it does, how you can access the feature and how it compares to OpenClaw. Keep reading to find out more about NemoClaw.
We’ve also put together a list of the best laptops and best gaming laptops too.
What is NemoClaw?
NemoClaw is an open source stack by Nvidia that adds privacy and security controls to OpenClaw and simplifies running OpenClaw’s always-on assistants. Currently available in early preview, NemoClaw enables anyone to run always-on, self-evolving agents “with one command”.
So, how does NemoClaw actually work? Nvidia explains that NemoClaw is the OpenClaw plugin for Nvidia OpenShell, and moves OpenClaw into a “sandboxed environment where every network request, file access, and inference call is governed by declarative policy”.
Essentially, NemoClaw utilises the Nvidia Agent Toolkit, which we’ll dive into more later, and install OpenShell to “enforce policy-based privacy and security guardrails” into OpenClaw. NemoClaw combines this security with open source models, such as Nvidia Nemotron, to make AI agent deployment simpler, more secure and more cost effective too.
It’s also worth noting that NemoClaw uses any coding agent, and can tap open models that are running locally on the user’s system. Agents can then use frontier models running in the cloud. This combination of local and cloud models promises to provide a foundation for agents to “develop and learn new skills to complete tasks according to defined privacy and security guardrails”.
We touched upon the Nvidia Agent Toolkit, as NemoClaw uses its features to simplify running OpenClaw always-on assistants.
Most notably for NemoClaw, the Nvidia Agent Toolkit includes Nemotron which is a family of open models, datasets and technologies for building agentic AI systems. According to Nvidia, Nemotron is designed for “advanced reasoning, coding, visual understanding, agentic tasks, safety, speech and informational retrieval”.
There’s also OpenShell which gives autonomous agents the required access while enforcing security, network and privacy guardrails.
NemoClaw vs OpenClaw?
It might sound as if NemoClaw is a competitor to OpenClaw, however this isn’t the case. Instead, NemoClaw works with OpenClaw.
During GTC, Nvidia’s founder and CEO Jensen Huang praised OpenClaw and explained it “opened the next frontier of AI to everyone” and “brings people closer to AI”. So, NemoClaw was developed not to replace, but to optimise OpenClaw and add data privacy and security to autonomous agents. This, according to Nvidia, provides the “missing infrastructure layer beneath claws to give them the access they need to be productive, while enforcing policy-based security, network and privacy guardrails.”
