jetson-mcp

by Zalmotek

171 downloads
Not rated
GitHub

About

A MCP server for using natural language to monitor and remote control a Nvidia Jetson board

Details

Author
Zalmotek
Downloads
171
Categories
Other

- Provides MCP tools via SSE transport on the network.
- get_jetson_hw_info: reads module/carrier board info from config file.
- get_jetson_sw_info: reads Jetpack version and Linux kernel version.
- Includes scripts for easy installation and systemd service setup.
- Requires firewall to allow incoming TCP on port 8000.

Setting up with Highlight

This MCP is not yet compatible with Highlight’s one-click setup. However, you can still use it with Highlight by following these steps:

  1. Download and install Highlight from highlightai.com/download
  2. Navigate to the plugins tab and select "Add Custom Plugin"
  3. Configure the plugin with the settings below
    Plugin Name jetson-mcp
    Command (node, npx, python, etc.)

    Please refer to the README for specific instructions on how to obtain API keys or other required environment variables.

  4. Enable "Start Automatically" if you want the plugin to start when Highlight launches

From the repository

Clone the repo on the Jetson, run ./install.sh to set up a Python virtual environment, then run sudo ./setup_service.sh to create a systemd service listening on port 8000. Start the service with sudo systemctl start jetson-mcp.service. From a remote MCP client, connect to the SSE endpoint at http://<jetson_ip>:8000/sse.

Claude Desktop / Cursor

Paste into your MCP client config file to install this server.

{
    "mcpServers": {
        "jetson-mcp": {
            "jetson-mcp": {
                "command": "python",
                "args": [
                    "app/main.py"
                ]
            }
        }
    }
}

McpServers

{
    "jetson-mcp": {
        "command": "python",
        "args": [
            "app/main.py"
        ]
    }
}

jetson-mcp

A MCP (Model Context Protocol) server for using natural language to monitor and remotely control a Nvidia Jetson board from clients on the same network.

<p align="center">
Nvidia Jetson
</p>

This project uses the FastMCP library to create the server.

MCP Server Connected:

<p align="center">
MCP Server Connected
</p>

Features

Provides MCP tools accessible by network clients using the SSE (Server-Sent Events) transport.
get_jetson_hw_info: Reads /etc/nv_boot_control.conf to identify module/carrier board info.
get_jetson_sw_info: Reads /etc/nv_tegra_release (for Jetpack version) and /proc/version (for Linux kernel version).
Includes scripts for easy installation and systemd service setup.

Setup and Installation (on the 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

Running the Server (on the Jetson)

The recommended way to run the server is as a background service managed by systemd.

1. (Optional) Find Jetson IP/Hostname:
You'll need the Jetson's IP address or hostname to connect from other devices. 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) configured to run the server as the user who invoked 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 Service:
    sudo systemctl status jetson-mcp.service
# Check logs for errors
sudo journalctl -u jetson-mcp.service -f

5. Firewall: Ensure your Jetson's firewall (if active, e.g., ufw) allows incoming connections on port 8000 (or your chosen port). Example for ufw:
    sudo ufw allow 8000/tcp

Running Manually (for testing)

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

```bash
source venv/bin/activate

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.