home / mcp / jetsonmcp server

JetsonMCP Server

MCP server for managing and controlling NVIDIA Jetson Nano Super systems via SSH - AI-powered system administration, Docker management, and edge computing optimization

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "ajeetraina-jetsonmcp": {
      "command": "python",
      "args": [
        "-m",
        "jetsonmcp.server"
      ],
      "env": {
        "JETSON_HOST": "192.168.1.100",
        "JETSON_PASSWORD": "your_password",
        "JETSON_USERNAME": "your_username",
        "JETSON_POWER_MODE": "0",
        "JETSON_SSH_KEY_PATH": "~/.ssh/id_rsa",
        "CUDA_VISIBLE_DEVICES": "0"
      }
    }
  }
}

JetsonMCP is an MCP server that connects AI assistants to NVIDIA Jetson Nano systems for edge computing management, AI workload optimization, and system administration. It enables natural-language-driven control over CUDA, JetPack, Docker, Kubernetes, and remote SSH-based administration to simplify deploying and managing AI workloads at the edge.

How to use

You interact with JetsonMCP through an MCP client to issue natural-language requests that translate into optimized commands on your Jetson Nano. Use it to deploy models, manage CUDA and JetPack, monitor hardware, optimize containers, and orchestrate edge deployments. Practical use cases include deploying a model for inference, checking GPU memory, switching power modes for battery operation, updating software stacks, and setting up a lightweight edge Kubernetes cluster. When you ask questions in natural language, JetsonMCP handles the underlying steps and returns actionable results.

How to install

Prerequisites: you need a Jetson Nano with a compatible JetPack (4.6+ recommended), SSH access, a suitable power supply, internet connectivity for package installation, and a client machine with Python and Git available.

Step 1. Install on your Jetson Nano or target host.

# 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 .

Step 2. Prepare your environment variables and configuration template.

# Copy configuration template
cp .env.example .env

# Edit with your Jetson's details
nano .env

Step 3. Start the MCP server (as a stdio server) by running the runtime entry shown in the configuration example.

# Typical command to run the server (shown in configuration example):
python -m jetsonmcp.server

Additional configuration and runtime details

JetsonMCP supports configuring the remote Jetson via an environment file. The example configuration includes host, SSH credentials, optional SSH key path, and advanced options like power mode and CUDA device selection. You can set a private Docker registry for edge deployments and select a GPU device for your workloads.

Example keys you may set in the configuration file include: JETSON_HOST, JETSON_USERNAME, JETSON_PASSWORD, JETSON_SSH_KEY_PATH, JETSON_POWER_MODE, CUDA_VISIBLE_DEVICES, DOCKER_REGISTRY.

Troubleshooting and notes

If you encounter SSH connection issues, verify that SSH is enabled on the Jetson Nano and that the host IP is reachable from your client. Ensure passwordless sudo is configured if the server needs elevated permissions. For GPU-related tasks, verify CUDA visibility and device selection. If the MCP server fails to start, check that you are using the correct Python environment and that the current directory contains the server module.

Configuration samples

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

Security considerations

Use SSH keys or strong passwords for remote access, limit SSH exposure to trusted networks, and enable audit logging for operations. Keep the JetPack and CUDA toolkits up to date and isolate critical services with appropriate user privileges and firewall rules.

Monitoring & observability

Track GPU utilization, memory usage, temperature, and power mode in real time. Monitor AI workload metrics such as inference latency, throughput, and resource utilization to ensure your edge deployments perform within expected parameters.

Use cases

Edge AI development, IoT and sensor networks, and industrial applications are supported. Deploy models, manage hardware, orchestrate edge containers, and maintain remote devices with secure, scalable workflows.

Available tools

manage_ai_workloads

Model deployment, inference optimization, and CUDA management for AI workloads on Jetson devices.

manage_jetpack

JetPack SDK installation, updates, component management, and version control.

manage_frameworks

Installation and optimization of AI frameworks such as TensorFlow, PyTorch, OpenCV, and TensorRT.

manage_hardware

Power modes, temperature monitoring, fan control, and GPIO-related tasks.

manage_performance

CPU/GPU governors, frequency scaling, and thermal management for optimized performance.

manage_containers

Docker management and NVIDIA runtime configuration for GPU-accelerated workloads.

manage_orchestration

Edge Kubernetes (K3s) deployment and orchestration for distributed AI workloads.

manage_registry

Private registry setup and multi-arch image management for edge deployments.

manage_system

Package management, updates, service control, and networking tasks.

manage_security

Firewall, SSH keys, user management, and system hardening.

manage_monitoring

System metrics, logging, alerting, and remote monitoring.