Skip to content

AaronGoldsmith/interactive_draw_mcp

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Interactive Drawing MCP

This project is an example of a Model Context Protocol (MCP) server with an interactive drawing interface. It demonstrates how to build an MCP extension that includes a separate UI window for drawing, using Tkinter as the graphical interface.

Features

  • Drawing Grid Interface:

    • Offers a 16x16 grid where each cell can be toggled between two states (filled or empty).
    • Provides a clear grid button to reset all cells.
  • Server Capabilities:

    • Start Drawing Session: Initializes a new session and launches the UI.
    • Toggle Cell: Toggle the color of individual cells via a command interface.
    • Get Grid State: Retrieve the current grid configuration as a text representation.
  • Persistence: Grid state is saved to a JSON file (board_state.json) to maintain consistency across sessions.

  • UI and State Synchronization:

    • Real-Time UI Update: A background thread continuously watches for updates to the grid state file and reflects changes visually.

Installation

Option 1: Install from Source

  1. Clone this repository:

    git clone https://github.com/AaronGoldsmith/interactive_draw_mcp
    cd interactive_draw_mcp
    
  2. Install the package:

    pip install -e .
    

Usage

As a Standalone MCP Server

Run the server to open the drawing window:

interactive-draw-mcp

With Goose

Option 1: Using Goose CLI (recommended)

Start Goose with your extension enabled:

# Local Development
goose session --with-extension "python -m interactive-draw-mcp"

Tools

  • Server Tools:
    • start_drawing_session(): Initializes a drawing session and opens the UI.
    • toggle_cell_color(row, col): Toggles the cell's color at specified row and col.
    • get_grid_state(): Provides a string representation of the grid's current state.

Resources and References

Architecture Overview

The architecture of this project demonstrates:

  1. MCP Server: Manages communication between the Goose AI and the drawing interface via well-defined tools.
  2. Interactive UI: Provides a visual interface for the grid.
  3. Integration with Goose AI: Tool calls from Goose AI reflect directly onto the grid state and vice versa.

Benefits

  • Separation of Responsibilities: Maintains clear separation between server logic and UI presentation.
  • Improved User Experience: Offers intuitive visual feedback alongside text-based command interface.
  • Adaptability: Can be extended to more sophisticated applications involving AI and interactive UIs.

About

MCP Server for collaborative drawing with Large Language Models

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Languages