Jetson Remote Monitor

🚀 jetson-mcp

A Model Context Protocol (MCP) server for monitoring and remotely controlling Nvidia Jetson boards via natural language, allowing clients on the same network to connect.

This project uses the FastMCP library to create the server.

✨ Features

• Provide MCP tools for web clients using Server-Sent Events (SSE) transmission.
• get_jetson_hw_info: Read /etc/nv_boot_control.conf to identify module/carrier board information.
• get_jetson_sw_info: Read /etc/nv_tegra_release for the Jetpack version and /proc/version for the Linux kernel version.
• Include scripts for easy installation and systemd service setup.

📦 Installation and Setup (on Jetson)

1. Clone the repository:

   git clone https://github.com/Zalmotek/jetson-mcp
   cd jetson-mcp
   

2. Run the installation script: This script creates a Python virtual environment (venv/) and installs dependencies from requirements.txt.

   chmod +x install.sh
   ./install.sh
   

🚀 Quick Start

It is recommended to run the server as a systemd background service.

1. (Optional) Find the Jetson's IP/hostname: You need the Jetson's IP address or hostname for other devices to connect. Use commands like ip addr or hostname -I.

2. Run the service setup script: This script creates and enables a systemd service file (/etc/systemd/system/jetson-mcp.service), configuring the server to run as the user who invokes the script, listening on port 8000.

   chmod +x setup_service.sh
   sudo ./setup_service.sh
   

3. Start the service:

   sudo systemctl start jetson-mcp.service
   

4. Verify the service status:

   sudo systemctl status jetson-mcp.service
   # Check the logs for errors
   sudo journalctl -u jetson-mcp.service -f
   

5. Firewall settings: Ensure that the Jetson's firewall (if active, e.g., ufw) allows incoming connections on port 8000. For example, for ufw:

   sudo ufw allow 8000/tcp
   

Manual Run (for testing)

The server is best run directly using the Python interpreter, which will call the mcp.run() method in the script:

source venv/bin/activate
# The script itself will now call mcp.run() with SSE, host, and port settings
python app/main.py

Connecting from a Remote Client

Once the server is running on the Jetson and accessible on the network (port 8000 allowed through the firewall):

1. Identify the server address: Find the Jetson's IP address (e.g., 192.168.1.105) or its hostname (e.g., jetson-nano.local) on your local network.
2. Configure the client: In your MCP client application (which could be a custom script, a UI like MCP Inspector, or perhaps Cursor/Claude with network endpoint support), configure it to connect to the network address of the MCP server.
   • The specific connection method depends on the client, but it usually involves specifying the URL of the SSE endpoint:
     • http://<jetson_ip_or_hostname>:8000/sse (common pattern for SSE)

Note: In Cursor's mcp.json file, network endpoint support is disabled by default. You need to enable the relevant settings to allow network connections to the model service.

Function List

Here are the main functions of the server:

• Hardware Information Query:
  • get_jetson_hw_info: Returns detailed hardware information about the Jetson, including the model and serial number.
• Software Information Query:
  • get_jetson_sw_info: Provides the currently installed software versions and related system information.

💻 Usage Examples

1. After installing and running the server, the client can obtain hardware information using the following command:

   curl http://localhost:8000/api/hardware
   

2. To obtain software information, you can run:

   curl http://localhost:8000/api/software
   

In this way, you can easily monitor and manage the status and information of your Jetson device.