cobusgreyling

Nemoclaw Runtime — Development skill for Claude Code

Development community

A secure, always-on runtime for autonomous agents — gateway, policy guardrails, OpenShell sandbox, and durable skills.

How to install Nemoclaw Runtime

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cobusgreyling/nemoclaw-runtime and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Nemoclaw Runtime does

A secure, always-on runtime for autonomous agents — gateway, policy guardrails, OpenShell sandbox, and durable skills. Dependency-free reference implementation.

Alternatives in Development

  • Claudia — A powerful GUI app and Toolkit for Claude Code - Create custom agents, manage interactive Claude Code sessions 21k ★
  • Claudes C Compiler — Claude Opus 4.6 wrote a dependency-free C compiler in Rust, with backends targeting x86 (64- and 32-bit), ARM 2.6k ★
  • Vibekit — A simple SDK for safely running Codex, Gemini CLI, and Claude Code in a secure sandbox 1.7k ★

README

![NemoClaw](assets/header.jpg)

◤ NemoClaw

**A secure, always-on runtime for autonomous agents.**

Run an agent inside a sandbox, behind a gateway that routes inference and enforces policy on *every* action.

`gateway` · `policy guardrails` · `OpenShell sandbox` · `routed inference` · `durable skills`

[![CI](https://github.com/cobusgreyling/nemoclaw-runtime/actions/workflows/ci.yml/badge.svg)](https://github.com/cobusgreyling/nemoclaw-runtime/actions/workflows/ci.yml) ![runtime](https://img.shields.io/badge/runtime-dependency--free-76B900) ![python](https://img.shields.io/badge/python-3.10%2B-76B900) ![status](https://img.shields.io/badge/status-early%20preview%20%2F%20alpha-blue) ![license](https://img.shields.io/badge/license-MIT-green)

![NemoClaw control plane](assets/dashboard.png)

**What this is.** An independent, **conceptual reference implementation** inspired by the NemoClaw blueprint, the secure agent runtime that runs *OpenClaw* (and the team-aware *Hermes* agent) inside an *OpenShell* sandbox, with a gateway that connects the agent to inference, integrations, and policy. It is **not** affiliated with or endorsed by NVIDIA. Early preview / alpha, built to make the architecture concrete and runnable, not for production.


The idea in one line

A language model can *propose* an action. NemoClaw decides whether it *happens*.

Most of the intelligence in an agent system does not live in the model. It lives in the orchestration wrapped around it, the routing, the guardrails, the sandbox, the memory of what worked. NemoClaw is that wrapper, made small enough to read in one sitting.

**The model proposes. The gateway disposes.**


Quick install

The blueprint advertises a hosted one-liner:

curl -fsSL https://www.nvidia.com/nemoclaw.sh | bash

This repo ships the honest, inspectable local equivalent, read it, then run it:

git clone https://github.com/cobusgreyling/nemoclaw-runtime.gi