Giters

4cecoder / sinkies

Repository from Github https://github.com4cecoder/sinkiesRepository from Github https://github.com4cecoder/sinkies

CursorClaude MCP

A central Model Control Panel (MCP) for bidirectional integration between Cursor IDE and Claude Desktop.

Overview

This project provides a central server (MCP) that facilitates real-time, bidirectional communication and synchronization between the Cursor IDE and the Claude Desktop application. It allows developers to leverage the strengths of both AI assistants in a parallel coding workflow.

The MCP handles:

  • Connecting to both Cursor and Claude via dedicated adapters.
  • Translating communication protocols between the two systems.
  • Synchronizing project state (files, edits) in real-time.
  • Managing potential conflicts arising from simultaneous edits.

Architecture

┌───────────────┐      ┌───────────────────┐      ┌─────────────────┐
│               │      │                   │      │                 │
│  Cursor IDE   │◄────►│  Central MCP      │◄────►│  Claude Desktop │
│               │      │                   │      │                 │
└───────────────┘      └───────────────────┘      └─────────────────┘
                             │
                             ▼
                       ┌───────────────┐
                       │  Project State │
                       │  Sync Engine   │
                       └───────────────┘

Core Components

  1. MCP Server:
    • RESTful API and WebSocket server built with Express and Socket.IO.
    • Handles authentication and session management.
    • Manages overall state and communication flow.
  2. Client Adapters (CursorAdapter, ClaudeAdapter):
    • Interface with the specific APIs and protocols of Cursor and Claude.
    • Translate messages for consumption by the MCP and the respective AI.
  3. Synchronization Engine (SyncEngine, StateManager):
    • Monitors the target project directory for file changes using chokidar.
    • Manages project state versioning and uses diff for efficient updates.
    • Implements conflict resolution strategies.
    • Handles real-time synchronization over WebSockets.
  4. Context Processor (ContextProcessor):
    • Extracts and optimizes code context to be sent to the AI assistants, respecting token limits.

Key Technologies

  • Backend: Node.js, Express, Socket.IO
  • AI SDKs: @anthropic-ai/sdk, cursor-api-client (assumed), openai
  • Synchronization: chokidar, fs-extra, diff, simple-git, lockfile
  • Utilities: jsonwebtoken, winston/pino, dotenv, config, node-cache
  • Development: TypeScript (optional), Jest, Nodemon, ESLint, Prettier, PM2

Getting Started

(Instructions to be added: Installation, Configuration, Running the server)

Usage

(Instructions to be added: How to connect Cursor and Claude, basic workflow)

Contributing

(Guidelines to be added)

License

(License information to be added)

About