Genox AIIoT Platform

GENOX AIIoT

On-Device AI Agent Platform for the Physical World

Create, deploy, and manage custom AI agents that control computers, smartphones, IoT devices, drones, robots, and appliances — all powered by quantized local LLMs with edge-first architecture.

Edge-First Architecture
MCP Protocol
Quantized LLMs
No-Code Builder

Intelligence at the Edge

The next wave of AI isn't in the cloud — it's on every device you own. From a Raspberry Pi running a smart home agent to a drone performing real-time object detection, our platform brings AI to the physical world.

With 4-bit quantization, a 3B-parameter model runs in just 2GB of RAM. Combined with MCP (Model Context Protocol) for universal device communication, any hardware becomes an intelligent agent.

This is zero cloud dependency, zero monthly cost, zero data leaving your premises. Your AI stays local, your data stays private, and your agents stay fast.

Platform Characteristics

Inference Latency<1ms
on-device, no network round-trip
Minimum Agent RAM80MB
nano-class quantized model (Q4)
Supported Protocols10+
IoT, drone, robotics, smart home, and more
Monthly Infrastructure Cost$0
fully local execution

Universal Device Control

One platform to create AI agents for every device in your ecosystem

Computer & Mobile

GUI automation via screen capture + VLM grounding, code execution, browser control, and smartphone ADB/Shortcuts integration

IoT & Smart Home

Home Assistant + Matter/Zigbee/MQTT integration for lighting, HVAC, security cameras, and energy management

Drones & Robots

MAVLink/PX4 drone control with autonomous patrol, YOLOv8 edge detection, and robotic arm manipulation via MCP

Appliances & Vehicles

SmartThings/Tuya appliance APIs, OBD-II vehicle diagnostics, wearable health data, and EV charging management

Core Capabilities

No-Code Agent Builder

Visual Workflow Designer: Drag-and-drop interface for defining agent behaviors, triggers, and device interactions without writing a single line of code.

YAML Agent Schema: Declarative agent configuration covering model selection, device bindings, behavioral rules, and safety constraints.

→ Select device → Define goals → Set rules → Choose model → Deploy

Safety Guard System

Multi-Layer Verification: Rate limiting, permission checks, anomaly detection, and human-in-the-loop approval for high-risk actions.

Sandboxed Execution: Each agent runs in isolation with network firewalling, encrypted MCP communication (TLS 1.3), and immutable audit logs.

→ Rollback engine restores previous device state on any error

Quantization Pipeline

Automated Model Optimization: Convert open-source models to optimized edge formats for any hardware target — from microcontrollers to desktops.

Tiered Model System: From 80MB nano models for embedded devices to 8GB+ desktop models — the right model for every device class.

→ 4-bit quantization reduces model size by 65% with minimal quality loss

Skill Marketplace

Community Agent Skills: Browse, install, and share pre-built agent skills for smart home automation, drone patrol, health monitoring, pet care, and more.

Physical Device Skills: Unlike software-only skill systems, our marketplace includes skills for robotic arms, CNC machines, 3D printers, and industrial equipment.

→ One-click install with automatic device compatibility check

Differentiating Innovation

01
Ghost Mode

Record your manual device interactions — the AI analyzes the pattern and automatically generates an agent that replicates your workflow. Your morning routine becomes an agent in one recording session.

02
Agent DNA Evolution

Agent behavioral parameters evolve through genetic algorithms driven by your feedback. Over generations, your agents become perfectly tuned to your preferences — and optimal "DNA" can be shared on the marketplace.

03
Split-Brain Architecture

Hybrid local + remote inference: lightweight models handle routine tasks on-device while complex reasoning seamlessly offloads to larger cloud models. The best of both worlds with automatic routing.

System Architecture

From your desktop to every device — a unified AI agent stack

01

Application Layer

Agent Builder

Drag-and-drop workflow designer

Device Panel

Live status and control

Dashboard

Monitoring and analytics

02

Agent Runtime Engine

Task PlannerMemory ManagerTool RouterSafety GuardGhost RecorderDNA Evolution
03

Inference Engine

Quantized LLM RuntimeVision Model PipelineSpeech ProcessingMulti-format Optimizer
04

MCP Protocol Layer

Universal communication layer — translates AI agent commands into device-native protocols automatically

What You Can Control

One platform, every device class — with practical AI agents for real-world use

Smart Home & IoT
  • Automated climate control based on occupancy
  • Security camera monitoring with anomaly alerts
  • Energy usage optimization across all appliances
  • Voice-controlled lighting and scene management
Drones & UAVs
  • Autonomous patrol routes with obstacle avoidance
  • Real-time object detection during flight
  • Agricultural crop health surveillance
  • Delivery route planning and package drop-off
Robots & Industrial
  • Robotic arm manipulation via natural language
  • Quality inspection on production lines
  • Warehouse inventory tracking and sorting
  • CNC machine operation and 3D print management
Computers & Servers
  • GUI automation via screen capture and AI grounding
  • Automated code review and deployment pipelines
  • System health monitoring with predictive alerts
  • Data processing and report generation agents
Smartphones & Tablets
  • App automation through accessibility APIs
  • Personal schedule and task management agents
  • Health data aggregation from phone sensors
  • Smart notification filtering and summarization
Vehicles & Fleet
  • OBD-II diagnostics with predictive maintenance
  • Fleet tracking and route optimization
  • EV charging schedule management
  • Driver behavior analysis and safety alerts
Wearables & Health
  • Continuous health metric monitoring
  • Exercise pattern analysis and coaching
  • Sleep quality tracking with recommendations
  • Emergency fall detection and alert dispatch
Energy & Utility
  • Solar panel output optimization
  • Smart grid load balancing agents
  • Water usage monitoring and leak detection
  • Building energy audit and efficiency reports

Custom Integration

Need to connect a device or system we don't support yet? Our platform is built for extensibility. Customers can request custom MCP bridges, device drivers, and specialized AI agent skills tailored to their unique hardware and workflows.

1

Custom MCP protocol bridges for proprietary hardware

2

Industry-specific AI agent templates and workflows

3

Private model fine-tuning for your domain data

4

On-premise deployment with dedicated support

5

White-label solutions for OEM partners

Contact us at admin@genox.one for custom integration requirements

How Models Scale

The platform automatically selects the right AI model size for each target device — from a tiny model on a microcontroller to a powerful model on your workstation. No manual configuration needed.

Nano(< 500 MB)

Microcontrollers, sensors, and embedded boards

Edge(0.5 – 3 GB)

Raspberry Pi, Jetson, single-board computers

Desktop(3 – 8 GB)

Workstations, laptops, and gaming PCs

Server(8 GB+)

Cloud instances and training clusters

Development Status

Genox AIIoT is currently in active development. The core Electron shell, Vue.js frontend, and Python/FastAPI backend are established. We are building towards full llama.cpp integration, MCP protocol layer, and No-Code Agent Builder.

Our roadmap targets MVP with Computer Use and Home Assistant agents, followed by drone control, smartphone integration, and the Skill Marketplace.

Pre-Alpha Development
MVP Target Q2 2026