ai-agent

NemoClaw: Secure AI Agents

Name: NemoClaw: Secure AI Agents
Author: NVIDIA

Run AI agents more securely with NemoClaw. For startup founders, with 21k+ GitHub stars.

22,316 stars3,055 forksTypeScriptHealth Score 8/10Updated 8/30/2026100% free · open source

What it does

NemoClaw lets you run AI agents like Hermes and OpenClaw more securely inside NVIDIA OpenShell with managed inference, enhancing the security and efficiency of your AI workflows

Install / run

git clone https://github.com/NVIDIA/NemoClaw.git

When to use it

•When you need to deploy AI agents in a secure environment
•When using NVIDIA OpenShell for AI inference
•When working with Hermes or OpenClaw agents and requiring enhanced security features

Quick start

1Clone the NemoClaw repository using the command `git clone https://github.com/NVIDIA/NemoClaw.git`
2Navigate into the cloned directory with `cd NemoClaw`
3Create a new NVIDIA OpenShell environment by following the instructions in the NemoClaw README
4Modify the `config.json` file to specify your AI agent and inference settings as described in the README
5Run your AI agent using the managed inference functionality provided by NemoClaw, as outlined in the README's examples

Ready-to-paste prompt

To run an example AI agent with managed inference, use a command similar to `nemo_claw --agent hermes --model your_model.pt --infer` as specified in the NemoClaw documentation

Heads up: Ensure you have the necessary NVIDIA drivers and OpenShell installed on your system before attempting to use NemoClaw, as it relies on these components for secure and efficient AI agent deployment

Saves to your device

Topics

ai-agents

nvidia

openclaw

openshell

sandboxing

typescript

How NemoClaw: Secure AI Agents works

Codeflow

Free to inspect

NemoClaw: Secure AI Agents is a very large TypeScript project (~962k lines across 3287 code files, plus 1641 test files). It is a full software project: use it through its install path rather than reading it end to end. Last commit this month, Apache-2.0 license, has a test suite.

Size

Very large codebase

~962k lines · 3287 code files · days to read — use, don't read

Setup

Developer setup

A real software project. Use it via its install path; don't expect to read it all.

Runs on

Node.js · Docker

No API keys detected