NAVADA EDGE runs NemoClaw — NVIDIA's open-source agentic runtime powered by Nemotron 120B. A secure, always-on AI agent with kernel-level sandboxing, MCP tool integration, and .NET Aspire tooling under construction.
Three-node hybrid mesh with NemoClaw as the autonomous intelligence layer
Kernel-level isolation via Landlock + seccomp + netns. Every file access, network call and inference request is policy-governed.
Nemotron 3 Super 120B on NVIDIA's infrastructure. Free nvapi- key. No GPU required on device. 40 req/min.
Autonomous execution: email, Telegram, web browsing, code, lead pipeline, trading reports — 24 hours a day.
Custom tools being built in C# .NET Aspire. NemoClaw calls NAVADA business logic via Model Context Protocol.
Lee Akpareva MBA, MA is the founder of NAVADA — an AI strategy consultancy and UK-Africa venture capital initiative building at the intersection of autonomous systems, enterprise intelligence, and emerging market infrastructure.
With 17+ years of digital transformation experience across British Airways, DHL, Farfetch, and the NHS, Lee brings a rare combination of enterprise delivery experience and hands-on technical building. He codes production systems, fine-tunes language models, designs robotic enclosures in Blender, and deploys infrastructure at 11pm on a laptop in London.
NAVADA EDGE is the physical and digital embodiment of that philosophy — a three-node autonomous AI mesh that runs 24/7, managing leads, trading intelligence, news digests, and infrastructure monitoring without human intervention.
NemoClaw is the newest layer: a secured, sandboxed AI agent powered by NVIDIA Nemotron 120B, wired into the NAVADA stack and extensible via .NET MCP tools being built in Visual Studio.
Why this moment matters — and why NAVADA is building the infrastructure now
We talk about AI as a feature — a chatbot, a copilot, a tool. That framing is already obsolete. What is happening is a structural shift equivalent to the emergence of the operating system itself. Just as Windows and macOS became the layer through which humans interact with computing, AI agents are becoming the layer through which humans and organisations interact with everything — their data, workflows, markets, communications.
The organisations and individuals who understand this early won't just use AI — they will build the infrastructure that others run on. NAVADA EDGE is that infrastructure: a sovereign, autonomous, extensible AI mesh that belongs to one operator and does exactly what that operator needs, 24 hours a day.
OpenClaw is the first mass-scale realisation of this thesis. NVIDIA's NemoClaw adds enterprise-grade security: policy-governed sandboxing, kernel-level isolation, inference routing. For NAVADA, this means an always-on agent that can manage leads, execute trades, respond to emails, monitor infrastructure, and communicate via Telegram — without human intervention. Not a chatbot. An operator.
Running AI on your own infrastructure means your data, your models, your context — on your terms. Not rented from a SaaS provider who can revoke access.
The value is not in the answer — it is in the action. An agent that executes tasks end-to-end is worth 100x an assistant that merely responds.
MCP tools mean the agent learns your business logic. Every .NET Aspire tool built is a new capability the agent can deploy autonomously.
The organisations that adopt autonomous AI infrastructure earliest in emerging economies will define the next decade of competitive advantage.
NAVADA EDGE is not a demo. It runs real workloads, real pipelines, real infrastructure. NAVADA Agent is live in the chat below.
Animated diagram of the agent's position within the full infrastructure mesh
The secure, policy-governed runtime layer that makes OpenClaw enterprise-grade
NemoClaw is NVIDIA's open-source stack that wraps OpenClaw with enterprise-grade security, sandboxing, and policy controls. It runs your agent in an isolated Docker container governed by OpenShell — a kernel-level runtime enforcing what the agent can read, write, call on the network, and use for inference.
For NAVADA EDGE, NemoClaw is the autonomous brain pointed at .NET tools via MCP — executing business logic, querying databases, sending Telegram messages and emails, all with full policy enforcement.
You (Telegram / Browser / Terminal) ↓ OpenClaw Agent ← Nemotron 120B brain ↓ OpenShell Gateway ← security layer ↓ ┌──────────────────────────────────┐ │ navada Sandbox │ │ Landlock + seccomp + netns │ │ writes: /sandbox + /tmp only │ │ network: policy-filtered │ └──────────────────────────────────┘ ↓ MCP calls .NET MCP Tools ← NAVADA logic ↓ NVIDIA Cloud ← Nemotron 120B
Every step taken to get NemoClaw running on Windows 11 + WSL2 — including every battle
Linux commands fail in PowerShell. Enter WSL and navigate to ~ not /mnt/c.
Sandbox image is 1.17GB — OOM killed with default WSL memory. Fixed via .wslconfig from Admin PowerShell.
Ubuntu 24.04 requires cgroup host mode for Docker Desktop.
Windows process PID 5824 was holding 8080. Admin PowerShell + WinNAT restart required.
7-step wizard: gateway → sandbox build → API key → Nemotron → policies.
Steps 5+6 complete — inference locked to Nemotron 120B, OpenClaw gateway launched.
Connect via tmux. Gateway token: navada2025. Identity: NAVADA, sharp and direct.
C# tools in Visual Studio + .NET Aspire — NemoClaw calls them via Model Context Protocol
Install inside navada sandbox to expand autonomous capabilities
nemoclaw navada connect → openclaw tui --token navada2025 → paste GitHub skill URL into chatSingle HTML file. Zero build step. Live in under 5 minutes on Cloudflare Pages.
Cloudflare Pages requires the entry file to be called index.html
Upload directly — no GitHub, no build step, no config needed.
Domain is already on Cloudflare — auto-configures, no manual DNS needed.
Wire the chat widget to your live NemoClaw agent endpoint on Node B.
Three-node hybrid mesh · Tailscale WireGuard · Zero services exposed to internet