JetsonMCP

An MCP server that connects AI assistants to NVIDIA Jetson Nano Super systems for comprehensive edge computing management, AI workload optimization, and system administration.

JetsonMCP enables AI assistants like Claude to help configure and manage Jetson Nano systems through SSH connections. From AI model deployment to system optimization - ask questions in natural language instead of learning complex CUDA and Linux commands.

🚀 What Makes JetsonMCP Special

Built for Edge AI Computing

CUDA toolkit management and optimization
JetPack SDK integration and updates
AI framework installation (TensorFlow, PyTorch, TensorRT)
Model deployment and inference optimization

Hardware-Specific Optimizations

GPU memory management and monitoring
Power mode configuration (10W/5W modes)
Temperature monitoring with thermal throttling
Fan curve management and cooling optimization

Container & Orchestration

Docker container management for AI workloads
NVIDIA Container Toolkit integration
Kubernetes edge deployment support
Multi-architecture container support (ARM64)

🎯 How It Works

Natural language requests are translated through Claude Desktop via MCP protocol to optimized commands executed on your Jetson Nano.

Examples:

AI & ML Operations:

"Deploy YOLOv5 model for object detection" - Downloads, optimizes, and runs inference
"Check CUDA memory usage" - Monitors GPU utilization and memory allocation
"Switch to 5W power mode" - Optimizes power consumption for battery operation
"Install TensorRT optimization" - Sets up high-performance inference engine

System Management:

"Monitor GPU temperature while running inference" - Real-time thermal monitoring
"Update JetPack to latest version" - Manages NVIDIA software stack updates
"Optimize Docker for AI workloads" - Configures runtime for GPU acceleration

Edge Computing:

"Deploy lightweight Kubernetes cluster" - Sets up K3s for edge orchestration
"Configure remote model serving" - Sets up inference endpoints
"Monitor system resources during AI tasks" - Performance profiling and optimization

🔧 Key Features

Edge AI Management

CUDA Toolkit Integration - Automatic CUDA environment setup and management
JetPack Management - SDK updates, component installation, and version control
AI Framework Support - TensorFlow, PyTorch, OpenCV, TensorRT optimization
Model Deployment - Automated model conversion, optimization, and serving

Hardware Optimization

Power Management - Dynamic power mode switching (10W/5W/MAXN)
Thermal Management - Temperature monitoring with automatic throttling
GPU Monitoring - Memory usage, utilization, and performance metrics
Fan Control - Custom fan curves and cooling optimization

Container Orchestration

NVIDIA Docker - GPU-accelerated container runtime management
Edge Kubernetes - K3s deployment for distributed AI workloads
Multi-arch Support - ARM64 container management and deployment
Registry Management - Private registry setup for edge deployments

System Administration

Remote Management - SSH-based secure system administration
Package Management - APT and snap package installation/updates
Service Management - Systemd service control and monitoring
Backup & Recovery - System state management and restoration

📋 Prerequisites

Jetson Nano Setup

Fresh JetPack Installation (4.6+ recommended)
SSH Access Enabled
Adequate Power Supply (5V/4A recommended for full performance)
MicroSD Card (64GB+ recommended) or NVMe SSD
Internet Connectivity for package installation

Network Configuration

Static IP recommended for consistent access
Firewall configured to allow SSH (port 22)
Optional: VPN setup for remote access

⚡ Quick Start

1. Prepare Your Jetson Nano

# Enable SSH (if not already enabled)
sudo systemctl enable ssh
sudo systemctl start ssh

# Configure passwordless sudo (required for system management)
sudo visudo
# Add: yourusername ALL=(ALL) NOPASSWD:ALL

# Get your Jetson's IP address
hostname -I

2. Install JetsonMCP

# Clone repository
git clone https://github.com/ajeetraina/jetsonMCP.git
cd jetsonMCP

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -e .

3. Configure Connection

# Copy configuration template
cp .env.example .env

# Edit with your Jetson's details
nano .env

Required .env settings:

JETSON_HOST=192.168.1.100        # Your Jetson's IP address
JETSON_USERNAME=your_username     # SSH username
JETSON_PASSWORD=your_password     # SSH password (or use SSH key)
# JETSON_SSH_KEY_PATH=~/.ssh/id_rsa  # Alternative: SSH key path

# Optional: Advanced Configuration
JETSON_POWER_MODE=0              # 0=MAXN, 1=5W, 2=10W
CUDA_VISIBLE_DEVICES=0           # GPU device selection
DOCKER_REGISTRY=localhost:5000   # Private registry for edge deployments

4. Claude Desktop Integration

Add to your Claude Desktop configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "jetsonmcp": {
      "command": "python",
      "args": ["-m", "jetsonmcp.server"],
      "cwd": "/absolute/path/to/jetsonMCP"
    }
  }
}

Restart Claude Desktop to load the server.

🛠️ Available Tools

AI & ML Management

manage_ai_workloads - Model deployment, inference optimization, CUDA management
manage_jetpack - JetPack SDK installation, updates, component management
manage_frameworks - TensorFlow, PyTorch, OpenCV, TensorRT installation

Hardware Control

manage_hardware - Power modes, temperature monitoring, fan control, GPIO
manage_performance - CPU/GPU governors, frequency scaling, thermal management
manage_storage - SSD optimization, swap configuration, disk management

Container Operations

manage_containers - Docker management, NVIDIA runtime, GPU acceleration
manage_orchestration - Kubernetes/K3s deployment, edge computing setup
manage_registry - Private registry setup, multi-arch image management

System Administration

manage_system - Package management, updates, service control, networking
manage_security - Firewall, SSH keys, user management, system hardening
manage_monitoring - System metrics, logging, alerting, remote monitoring

🔧 Advanced Configuration

Power Management

# Available power modes
# 0: MAXN (15W) - Maximum performance
# 1: 5W - Power efficient mode  
# 2: 10W - Balanced mode
sudo nvpmodel -m 1  # Switch to 5W mode

CUDA Environment

# Verify CUDA installation
nvcc --version
nvidia-smi

# Set CUDA paths (automatically configured by JetsonMCP)
export PATH=/usr/local/cuda/bin:$PATH
export LDFLAGS=-L/usr/local/cuda/lib64:$LDFLAGS

Docker GPU Support

# NVIDIA Container Toolkit (managed by JetsonMCP)
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

🧪 Testing & Development

Run Tests

# Run all tests
make test

# Specific component tests
make test-ai          # AI workload management
make test-hardware    # Hardware control
make test-containers  # Container operations
make test-system      # System administration

# Integration tests
python run_tests.py --integration --coverage

Development Setup

# Install development dependencies
pip install -e ".[dev]"

# Pre-commit hooks
pre-commit install

# Code formatting
black jetsonmcp/
isort jetsonmcp/

# Type checking
mypy jetsonmcp/

📊 Monitoring & Observability

System Metrics

CPU/GPU Utilization - Real-time performance monitoring
Memory Usage - RAM and GPU memory tracking
Temperature Sensors - Thermal monitoring and alerts
Power Consumption - Current power mode and usage metrics

AI Workload Metrics

Inference Latency - Model performance benchmarking
Throughput - Requests per second for deployed models
Resource Utilization - GPU memory and compute efficiency
Model Accuracy - Performance validation and monitoring

🔒 Security Features

SSH Security

Host key verification and rotation
Connection timeouts and retry logic
Credential management and cleanup
Audit logging for all operations

Container Security

Image vulnerability scanning
Runtime security policies
Network isolation and segmentation
Secrets management for AI models

System Hardening

Firewall configuration management
User privilege separation
System update automation
Security patch monitoring

🚀 Use Cases

Edge AI Development

Rapid prototyping of AI applications
Model optimization and benchmarking
Distributed inference deployment
Real-time computer vision applications

IoT & Sensor Networks

Sensor data processing and analysis
Edge computing orchestration
Remote device management
Predictive maintenance systems

Industrial Applications

Quality control and inspection
Predictive analytics
Autonomous systems development
Industrial IoT integration

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Development Priorities

AI Framework Integration - Support for additional ML frameworks
Edge Orchestration - Advanced Kubernetes edge deployment
Hardware Abstraction - Support for other Jetson platforms (AGX, Xavier)
Monitoring Enhancement - Advanced telemetry and observability

JetsonMCP

An MCP server that connects AI assistants to NVIDIA Jetson Nano Super systems for comprehensive edge computing management, AI workload optimization, and system administration.

🚀 What Makes JetsonMCP Special

Built for Edge AI Computing

CUDA toolkit management and optimization
JetPack SDK integration and updates
AI framework installation (TensorFlow, PyTorch, TensorRT)
Model deployment and inference optimization

Hardware-Specific Optimizations

GPU memory management and monitoring
Power mode configuration (10W/5W modes)
Temperature monitoring with thermal throttling
Fan curve management and cooling optimization

Container & Orchestration

Docker container management for AI workloads
NVIDIA Container Toolkit integration
Kubernetes edge deployment support
Multi-architecture container support (ARM64)

🎯 How It Works

Natural language requests are translated through Claude Desktop via MCP protocol to optimized commands executed on your Jetson Nano.

Examples:

AI & ML Operations:

"Deploy YOLOv5 model for object detection" - Downloads, optimizes, and runs inference
"Check CUDA memory usage" - Monitors GPU utilization and memory allocation
"Switch to 5W power mode" - Optimizes power consumption for battery operation
"Install TensorRT optimization" - Sets up high-performance inference engine

System Management:

"Monitor GPU temperature while running inference" - Real-time thermal monitoring
"Update JetPack to latest version" - Manages NVIDIA software stack updates
"Optimize Docker for AI workloads" - Configures runtime for GPU acceleration

Edge Computing:

"Deploy lightweight Kubernetes cluster" - Sets up K3s for edge orchestration
"Configure remote model serving" - Sets up inference endpoints
"Monitor system resources during AI tasks" - Performance profiling and optimization

🔧 Key Features

Edge AI Management

CUDA Toolkit Integration - Automatic CUDA environment setup and management
JetPack Management - SDK updates, component installation, and version control
AI Framework Support - TensorFlow, PyTorch, OpenCV, TensorRT optimization
Model Deployment - Automated model conversion, optimization, and serving

Hardware Optimization

Power Management - Dynamic power mode switching (10W/5W/MAXN)
Thermal Management - Temperature monitoring with automatic throttling
GPU Monitoring - Memory usage, utilization, and performance metrics
Fan Control - Custom fan curves and cooling optimization

Container Orchestration

NVIDIA Docker - GPU-accelerated container runtime management
Edge Kubernetes - K3s deployment for distributed AI workloads
Multi-arch Support - ARM64 container management and deployment
Registry Management - Private registry setup for edge deployments

System Administration

Remote Management - SSH-based secure system administration
Package Management - APT and snap package installation/updates
Service Management - Systemd service control and monitoring
Backup & Recovery - System state management and restoration

📋 Prerequisites

Jetson Nano Setup

Fresh JetPack Installation (4.6+ recommended)
SSH Access Enabled
Adequate Power Supply (5V/4A recommended for full performance)
MicroSD Card (64GB+ recommended) or NVMe SSD
Internet Connectivity for package installation

Network Configuration

Static IP recommended for consistent access
Firewall configured to allow SSH (port 22)
Optional: VPN setup for remote access

⚡ Quick Start

1. Prepare Your Jetson Nano

# Enable SSH (if not already enabled)
sudo systemctl enable ssh
sudo systemctl start ssh

# Configure passwordless sudo (required for system management)
sudo visudo
# Add: yourusername ALL=(ALL) NOPASSWD:ALL

# Get your Jetson's IP address
hostname -I

2. Install JetsonMCP

# Clone repository
git clone https://github.com/ajeetraina/jetsonMCP.git
cd jetsonMCP

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -e .

3. Configure Connection

# Copy configuration template
cp .env.example .env

# Edit with your Jetson's details
nano .env

Required .env settings:

JETSON_HOST=192.168.1.100        # Your Jetson's IP address
JETSON_USERNAME=your_username     # SSH username
JETSON_PASSWORD=your_password     # SSH password (or use SSH key)
# JETSON_SSH_KEY_PATH=~/.ssh/id_rsa  # Alternative: SSH key path

# Optional: Advanced Configuration
JETSON_POWER_MODE=0              # 0=MAXN, 1=5W, 2=10W
CUDA_VISIBLE_DEVICES=0           # GPU device selection
DOCKER_REGISTRY=localhost:5000   # Private registry for edge deployments

4. Claude Desktop Integration

Add to your Claude Desktop configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "jetsonmcp": {
      "command": "python",
      "args": ["-m", "jetsonmcp.server"],
      "cwd": "/absolute/path/to/jetsonMCP"
    }
  }
}

Restart Claude Desktop to load the server.

🛠️ Available Tools

AI & ML Management

manage_ai_workloads - Model deployment, inference optimization, CUDA management
manage_jetpack - JetPack SDK installation, updates, component management
manage_frameworks - TensorFlow, PyTorch, OpenCV, TensorRT installation

Hardware Control

manage_hardware - Power modes, temperature monitoring, fan control, GPIO
manage_performance - CPU/GPU governors, frequency scaling, thermal management
manage_storage - SSD optimization, swap configuration, disk management

Container Operations

manage_containers - Docker management, NVIDIA runtime, GPU acceleration
manage_orchestration - Kubernetes/K3s deployment, edge computing setup
manage_registry - Private registry setup, multi-arch image management

System Administration

manage_system - Package management, updates, service control, networking
manage_security - Firewall, SSH keys, user management, system hardening
manage_monitoring - System metrics, logging, alerting, remote monitoring

🔧 Advanced Configuration

Power Management

# Available power modes
# 0: MAXN (15W) - Maximum performance
# 1: 5W - Power efficient mode  
# 2: 10W - Balanced mode
sudo nvpmodel -m 1  # Switch to 5W mode

CUDA Environment

# Verify CUDA installation
nvcc --version
nvidia-smi

# Set CUDA paths (automatically configured by JetsonMCP)
export PATH=/usr/local/cuda/bin:$PATH
export LDFLAGS=-L/usr/local/cuda/lib64:$LDFLAGS

Docker GPU Support

# NVIDIA Container Toolkit (managed by JetsonMCP)
distribution=$(. /etc/os-release;echo $ID$VERSION_ID)
curl -s -L https://nvidia.github.io/nvidia-docker/gpgkey | sudo apt-key add -
curl -s -L https://nvidia.github.io/nvidia-docker/$distribution/nvidia-docker.list | sudo tee /etc/apt/sources.list.d/nvidia-docker.list

🧪 Testing & Development

Run Tests

# Run all tests
make test

# Specific component tests
make test-ai          # AI workload management
make test-hardware    # Hardware control
make test-containers  # Container operations
make test-system      # System administration

# Integration tests
python run_tests.py --integration --coverage

Development Setup

# Install development dependencies
pip install -e ".[dev]"

# Pre-commit hooks
pre-commit install

# Code formatting
black jetsonmcp/
isort jetsonmcp/

# Type checking
mypy jetsonmcp/

📊 Monitoring & Observability

System Metrics

CPU/GPU Utilization - Real-time performance monitoring
Memory Usage - RAM and GPU memory tracking
Temperature Sensors - Thermal monitoring and alerts
Power Consumption - Current power mode and usage metrics

AI Workload Metrics

Inference Latency - Model performance benchmarking
Throughput - Requests per second for deployed models
Resource Utilization - GPU memory and compute efficiency
Model Accuracy - Performance validation and monitoring

🔒 Security Features

SSH Security

Host key verification and rotation
Connection timeouts and retry logic
Credential management and cleanup
Audit logging for all operations

Container Security

Image vulnerability scanning
Runtime security policies
Network isolation and segmentation
Secrets management for AI models

System Hardening

Firewall configuration management
User privilege separation
System update automation
Security patch monitoring

🚀 Use Cases

Edge AI Development

Rapid prototyping of AI applications
Model optimization and benchmarking
Distributed inference deployment
Real-time computer vision applications

IoT & Sensor Networks

Sensor data processing and analysis
Edge computing orchestration
Remote device management
Predictive maintenance systems

Industrial Applications

Quality control and inspection
Predictive analytics
Autonomous systems development
Industrial IoT integration

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Development Priorities

AI Framework Integration - Support for additional ML frameworks
Edge Orchestration - Advanced Kubernetes edge deployment
Hardware Abstraction - Support for other Jetson platforms (AGX, Xavier)
Monitoring Enhancement - Advanced telemetry and observability