NemoClaw

Why we included this project

Teams that already run OpenClaw, Hermes, or LangChain Deep Agents often want a way to keep those agents from reaching too far into the host machine. NemoClaw is a reference stack that runs them inside NVIDIA OpenShell sandboxes and adds the operational pieces that are easy to get wrong: managed inference, network egress policy, snapshots, and lifecycle operations through a CLI. On a DGX or WSL host, the installer guides you through setup, defaults to OpenClaw, and can pick an inference provider and model for you. It is not a standalone agent, so it only helps if you already use one of the supported runtimes. The project is early-stage alpha, so plan for active iteration and review the security documentation before trusting it with sensitive workloads.