MCP Enhanced Feedback Server (mcp-feedback-enhanced)

Project Overview

mcp-feedback-enhanced is an enhanced MCP (Model Context Protocol) server specifically designed for interactive user feedback and command execution within AI-assisted development tools. This project supports a dual-interface mode (Web UI and desktop application), featuring intelligent environment detection and cross-platform compatibility.

Core Features

• Dual Interface Support: Supports both Web UI and native desktop applications.
• Smart Environment Detection: Automatically detects the running environment and selects the most suitable interface mode.
• Cross-Platform Compatibility: Supports Windows, macOS, and Linux systems.
• Cost Optimization: By guiding the AI to confirm with the user instead of performing speculative operations, it can consolidate up to 25 tool calls into a single feedback-driven request, significantly reducing platform costs.

Supported Platforms

Supported AI development platforms include: Cursor, Cline, Windsurf

How It Works

The project's workflow is as follows:

1. AI Invocation → mcp-feedback-enhanced
2. Environment Detection → Automatically selects the appropriate interface
3. User Interaction → Command execution, text feedback, image upload
4. Feedback Delivery → Information returned to the AI
5. Process Continuation → Adjusts or concludes based on feedback

Key Features

Interface Modes

Qt GUI Interface

• Native experience in local environments
• Modular refactoring design
• Suitable for local development environments

Web UI Interface

• Modern interface design, suitable for remote SSH environments
• New architectural design
• Responsive layout, adapts to different screen sizes

Core Functionality

Image Processing Features

• Format Support: PNG, JPG, JPEG, GIF, BMP, WebP
• Upload Methods: Drag & drop files + clipboard paste (Ctrl+V)
• Automatic Processing: Smart compression ensures compliance with the 1MB limit

Multi-Language Support

• Three Languages: English, Traditional Chinese, Simplified Chinese
• Smart Detection: Automatically selects based on system language
• Real-time Switching: Change language directly within the interface

Shortcut Key Support

• Ctrl+Enter: Submit feedback (supports main keyboard and numeric keypad)
• Ctrl+V: Directly paste clipboard images

Installation and Configuration

Quick Installation

# Install uv (if not already installed)
pip install uv

# Quick test
uvx mcp-feedback-enhanced@latest test

Basic Configuration

Add the following configuration to your AI assistant (e.g., Cursor):

{
  "mcpServers": {
    "mcp-feedback-enhanced": {
      "command": "uvx",
      "args": ["mcp-feedback-enhanced@latest"],
      "timeout": 600,
      "autoApprove": ["interactive_feedback"]
    }
  }
}

Advanced Configuration

{
  "mcpServers": {
    "mcp-feedback-enhanced": {
      "command": "uvx",
      "args": ["mcp-feedback-enhanced@latest"],
      "timeout": 600,
      "env": {
        "FORCE_WEB": "true",
        "MCP_DEBUG": "false"
      },
      "autoApprove": ["interactive_feedback"]
    }
  }
}

Environment Variable Configuration

Variable Name	Purpose	Values	Default Value
FORCE_WEB	Force Web UI	true/false	false
MCP_DEBUG	Debug mode	true/false	false
INCLUDE_BASE64_DETAIL	Full Base64 for images	true/false	false

Usage Recommendations

For best results, it is recommended to add the following rules to your AI assistant:

# MCP Interactive Feedback Rules
1. In any process, task, or conversation, whether inquiring, responding, or completing stage tasks, MCP mcp-feedback-enhanced must be invoked.
2. When user feedback is received, if the feedback content is not empty, MCP mcp-feedback-enhanced must be invoked again, and behavior adjusted based on the feedback.
3. Only when the user explicitly states "end" or "no more interaction needed" should the invocation of MCP mcp-feedback-enhanced stop, and then the process completes.
4. Unless an end command is received, all steps must repeatedly invoke MCP mcp-feedback-enhanced.

Testing Methods

Version Check

uvx mcp-feedback-enhanced@latest version

Interface-Specific Tests

# Qt GUI quick test
uvx mcp-feedback-enhanced@latest test --gui

# Web UI test (automatically runs continuously)
uvx mcp-feedback-enhanced@latest test --web

# Debug mode
MCP_DEBUG=true uvx mcp-feedback-enhanced@latest test

Local Development Testing

# Clone the project
git clone https://github.com/Minidoracat/mcp-feedback-enhanced.git
cd mcp-feedback-enhanced
uv sync

# Method 1: Standard test (recommended)
uv run python -m mcp_feedback_enhanced test

# Method 2: Full test suite
uvx --with-editable . mcp-feedback-enhanced test

# Method 3: Interface-specific tests
uvx --with-editable . mcp-feedback-enhanced test --gui  # Qt GUI test
uvx --with-editable . mcp-feedback-enhanced test --web  # Web UI test

Version History and Major Updates

Latest Features (v2.6.0 and above)

• Complete Refactoring: GUI and Web UI adopt a modular architecture
• Centralized Management: Reorganized folder structure for improved maintainability
• Interface Optimization: Modernized design and improved user experience
• macOS Interface Optimization: Specific improvements for macOS user experience
• Feature Enhancements: New settings options and automatic page closing function

Core Improvements

• Language Switching: Fixed content update issues when switching languages in Web UI
• About Page: Added an About page containing version information, project links, and acknowledgements
• Enhanced Hotkeys: Ctrl+Enter now supports numeric keypad
• Smart Image Paste: Ctrl+V directly pastes clipboard images
• Multi-language Architecture Refactoring: Dynamic loading and modular organization of language files

Frequently Asked Questions

Q: Encountering "Unexpected token 'D'" error

A: Debug output interference. Set MCP_DEBUG=false or remove the environment variable.

Q: Chinese characters are garbled

A: Fixed in v2.0.3. Update to the latest version: uvx mcp-feedback-enhanced@latest

Q: Image upload failed

A: Check file size (≤1MB) and format (PNG/JPG/GIF/BMP/WebP).

Q: Web UI fails to start

A: Set FORCE_WEB=true or check firewall settings.

Q: Gemini Pro 2.5 cannot parse images

A: Known issue. Gemini Pro 2.5 may not correctly parse uploaded image content. Tests show that Claude-4-Sonnet can analyze images correctly. It is recommended to use Claude models for better image understanding capabilities.

Project Value

The core value of this project lies in solving a critical problem in AI-assisted development: avoiding expensive speculative operations by AI. By establishing a human-AI interactive feedback loop, users can provide confirmation and guidance before the AI executes complex or costly operations, thereby:

1. Reducing Costs: Minimizing unnecessary API calls
2. Increasing Efficiency: Avoiding erroneous operational paths
3. Enhancing Control: Users maintain leadership over the development process
4. Improving Experience: A smoother human-AI collaborative development workflow