NemoClaw & DeepAgents

Interactive lessons for understanding NVIDIA's NemoClaw, OpenShell sandboxing, and LangChain's DeepAgents architecture. Start with the fundamentals, then dive into technical implementation details.

Lessons

Lesson 01

Fundamentals

What is NemoClaw, OpenShell, OpenClaw, Blueprint, and DeepAgents?

Lesson 02

Architecture Deep Dive

How the five security layers connect and enforce policies.

Lesson 03

dcode vs Claude Code

Comparing LangChain's dcode with Anthropic's Claude Code.

Lesson 04

Additional Questions

Deep questions an AI engineer should ask about these systems.

Lesson 05

DeepAgents Technical Deep Dive

Planning, memory, subagents, and HITL implementation details.

Lesson 06

Sandbox Technical Implementation

Landlock LSM, seccomp-bpf, namespaces, and kernel features.

Lesson 07

Blueprint to Agent Flow

The complete onboarding sequence from YAML to running agent.

Lesson 08

Human-in-the-Loop Deep Dive

Approval workflows, thresholds, and safety patterns.

Lesson 09

Multi-Agent Systems

Orchestration patterns, delegation, and coordination.

Lesson 10

Five Protection Layers Deep Dive

Technical deep dive on network, filesystem, process, gateway, inference.

Lesson 11

Harness Tuning & Policy Gateways

Domain-specific accuracy via harness tuning. Prompt-level vs enforcement-level rules. Sandbox boundary decisions.

Lesson 12

Sandbox Deployment Options & Kubernetes

Four deployment options: cloud providers, custom K8s, self-hosted OpenShell, hybrid. Helm charts, compute drivers, security comparison.

Reference

Glossary

25+ terms defined with sources.

Architecture Diagrams

Visual reference for system architecture.

DeepAgents Code Reference

Code snippets and API examples.

OpenShell Technical Internals

Kernel-level implementation details.

Sources