mycobot

This MCP server enables simple pick-and-place control of a myCobot robot. It relies on image feedback for object detection and requires an MCP client that can process and display image data. The server integrates with the DDS API to perform object detection and uses a settings.json file to configure the myCobot hardware and camera. To run, start the uvx-based MCP server with the provided settings path and DDS API token, then connect a compatible MCP client that can render image feedback and issue pick-and-place commands to the robot. The server exposes basic capabilities to move the arm, control the suction gripper, and specify pick and drop poses defined in your settings.

Typical workflow: the client requests an object to pick, the server processes the camera feed with DDS to detect the object, and once detected, it guides the myCobot through a sequence of poses to pick up and place the object at the defined drop location. The included sample settings.json outlines several named poses (home, capture, drop) and hardware parameters you can customize for your setup.

Prerequisites:

• A host machine with Node.js not required for this MCP (this server uses uv/uvx).
• Access to the DDS API token for object detection.
• Optional: Claude-Desktop on Ubuntu if you prefer a local GUI workflow.

Installation steps:

1. Install the runtime tooling (uv/uvx):

   • If using the uv/uvx approach, install the uv runtime as per your platform's instructions. Example for Ubuntu (adjust according to your environment):
     • sudo apt-get update
     • sudo apt-get install python3-pip # if Python tools are used in your setup
     • pip3 install uv # or follow the official uvx installation guide if applicable

2. Obtain an API token for object detection (DDS):

   • Register and issue an API key at the DDS service: https://cloud.deepdataspace.com/
   • Keep the token handy for the environment variable DDS_API_TOKEN.

3. Clone or obtain the MCP server repository:

   • git clone https://github.com/neka-nat/mycobot-mcp.git
   • cd mycobot-mcp

4. Prepare settings.json:

   • Create or edit data/settings.json (or the path you will point to with --settings-path).
   • Example structure is provided in the README under Parameters. Adjust camera_id, calibration paths, and myCobot settings (urdf_path, port, baud, etc.).

5. Run the MCP server via uvx:

   • Ensure the settings.json path is correct and that the environment variable DDS_API_TOKEN is set.
   • Example command: uvx mycobot-mcp --settings-path /path/to/settings.json

6. (Optional) If you prefer the Docker or npx/other routes, adapt accordingly once you know your deployment target.

Prerequisites notes:

• Make sure the user has permission to access the serial port (e.g., chmod 666 /dev/ttyACM0 or appropriate udev rules).
• Ensure the camera is accessible to the DDS service and that the image feedback pipeline is functioning before attempting pick-and-place tasks.

Tips and caveats:

• Image feedback is required for this MCP server to function correctly. Ensure your camera and DDS API token have proper access and that the client can render or pass through image frames.
• The sample settings.json includes hardware parameters (urdf_path, port, baud, end_effector_name, and motion presets). Update these values to reflect your actual hardware configuration.
• If using Claude-Desktop on Ubuntu, update claude_desktop_config.json with the mycobot settings and the correct path to settings.json.
• When networked with DDS, secure your API tokens and avoid exposing them in public configurations.
• If you encounter permission issues on serial ports, configure appropriate udev rules instead of using broad chmod in production.
• The MCP relies on predefined places: home, capture (camera capture position), and drop. You can expand these in settings.json to suit additional tasks.