NemoClaw

What NemoClaw does

NVIDIA NemoClaw provides guided onboarding, managed inference, network policy, managed integrations, OpenShell-backed state persistence and lifecycle operations through the NemoClaw CLI and its agent-specific aliases. Its goal is to run supported agents more safely inside NVIDIA OpenShell sandboxes.

Supported agents are OpenClaw (the default), Hermes and LangChain Deep Agents Code. On a supported DGX or Windows Subsystem for Linux host, an express install preset installs OpenClaw by default, so answer n at the prompt to choose Hermes, a sandbox name, an inference provider and a model interactively. Hermes has its own quickstart in the NVIDIA docs. When connecting to a Hermes sandbox from a light terminal, NemoClaw may install a managed nemoclaw-light Hermes skin for readable text, removes it again when no longer needed and keeps any skin you selected yourself.

Key features

• Runs agents inside NVIDIA OpenShell sandboxes
• Guided onboarding with a choice of agent, sandbox name, inference provider and model
• Managed inference and network policy
• Managed integrations and OpenShell-backed state persistence
• Lifecycle operations through the NemoClaw CLI and agent-specific aliases
• Dedicated Hermes quickstart in the documentation

When to use it

• Running a Hermes sandbox with controlled network access and managed inference
• Trying OpenClaw, Hermes or LangChain Deep Agents Code under one sandboxed stack
• Letting a coding agent guide the install with the official starter prompt

Who it is for: Teams and individuals who want to run Hermes or other agents inside NVIDIA OpenShell sandboxes with managed inference and network policy.

How it fits with Hermes Agent

Hermes is one of three supported agents, next to the default OpenClaw, and has its own Quickstart with Hermes guide in the NemoClaw documentation.

FAQ