…
Course home

Securing Agents with OpenShell and NemoClaw

Abstract

Modern agents are everywhere, and they’re conceptually simple: a model wired to tools, memory, and a routing decision that keeps running until the task is done. This course starts by building such a system from scratch, then connects those ideas to modern frameworks used by software engineers and non‑technical users alike. You’ll go from a single API call to agent coordination, grounded retrieval, deep planning, and safe deployment using NVIDIA OpenClaw and NemoClaw™.

1Confirm the model routes below. Chat and embeddings use models from build.nvidia.com by default. You can change each route's model and endpoint to suit your performance needs.
2Follow the modules in order for the default path, or jump to the task you need. Every page ends in a live browser artifact, with its JavaScript available for inspection and adaptation.
3Create a NemoClaw launchable to hosts the course's OpenClaw-in-OpenShell stack from the NemoClaw blueprint. This can be reproduced locally without GPU access using model endpoints.

Learning Objectives

Upon completion of this course, students will be able to:

  • Build a basic agent loop and identify its core components.
  • Implement reliable tool use and function calling within an agent system.
  • Design and coordinate multi-agent systems using structured routing patterns.
  • Utilize OpenShell to configure agent identities and ensure safe, sandboxed operations.
  • Deploy and manage autonomous agents while building persistent skill libraries.
A NemoClaw mascot holding the OpenShell shield beside a local AI workstation.
NemoClaw for OpenClaw in a local workstation setting.
Source: NVIDIA Build blueprint card.

Entering Your API Key

Your browser saves the chat and embedding routes across lessons and sends each key only to its selected endpoint. Use a custom route only when you have a reachable HTTPS endpoint.

Exploring the Material

Use the modules as a guided order; each page also remains useful on its own. Following the full path lets each module reuse its predecessor—the browser-hosted loop becomes a scoped workflow, that workflow moves into a persistent agent runtime, and the runtime's wider reach motivates the operating-system boundary that contains it. Start with the task you need. Every page ends in a live artifact and keeps the supporting code available. You can modify the examples in the browser, self-host them, or give this repository to a coding agent as context for a local deployment or application.

Module 1 · ~1 hour · 3 pages

Agent Basics · Loop · ReAct · Tools

Build the smallest agent loop, add ReAct reasoning, and turn plain functions into tool calls the model can request.
1a Loop · 1b ReAct · 1c Tools
Module 2 · ~1 hour · 3 pages

Advanced Patterns · Workflows · RAG · Deep Agents

Route work across calls, retrieve course material as a tool, and use planners that spawn sub-agents.
2a Workflows · 2b RAG · 2c Deep Agents
Module 3 · ~1 hour · 3 pages

NemoClaw Setup · OpenClaw · Always-On

Connect the NemoClaw launchable. Explore OpenClaw's gateway, workspace, skills, and scheduled runs inside OpenShell.
3a Kickstart · 3b OpenClaw · 3c Always-On
Module 4 · ~1 hour · 3 pages

OpenShell Safety · Modern CLIs · Going Further

Shift from OpenClaw behavior to the OpenShell policy boundary that NemoClaw assembles around it, then compare modern CLI agents and current safety patterns.
4a OpenShell · 4b Modern CLIs · 4c Going Further

What this course runs on

The course uses these three systems from the browser. Each card shows what the system contributes and opens the relevant service or catalog.

Models
build.nvidia.com
Use NVIDIA-hosted Nemotron models through an OpenAI-compatible API. Calls use your nvapi- key and the selected model route.
Search
Materials catalog
webSearch ranks the course glossary, papers, and posts. instantAnswer returns one matching glossary definition. Open the catalog to inspect the sources and search request.
Agent stack
NemoClaw launchable on Brev
NemoClaw assembles OpenClaw inside an OpenShell sandbox. The later modules connect to this Brev launchable. A local deployment uses the same components and can be adapted from the source.
This course is self-contained and runs entirely in GitHub Pages. Bring an nvapi- key and work through the four modules in order or as preferred.