# 🔍 AI Content Detector MCP Server

Detect AI-generated content, analyze writing style, check plagiarism, and humanize text. Recognizes 24 AI writing patterns. Sister product of HumanizerMCP - forms a "detect + humanize" closed loop.

## Features

- **8 MCP Tools**: detect_ai_content, analyze_writing_style, check_plagiarism, get_detection_report, humanize_content, batch_detect, compare_versions, get_ai_probability
- **24 AI Patterns**: Based on Wikipedia "Signs of AI writing" guide
- **Style Analysis**: Sentence rhythm, vocabulary diversity, personal voice, emotional richness
- **Plagiarism Check**: N-gram similarity and self-repetition detection
- **Humanization**: Actionable suggestions to reduce AI traces

## Quick Start

```bash
# Install
pip install -r requirements.txt

# Run (stdio mode)
python server.py

# Run (HTTP mode)
python server.py --transport http --port 8010
```

## Tools

| Tool | Description |
|------|-------------|
| `detect_ai_content` | Detect AI patterns with severity scores |
| `analyze_writing_style` | Analyze sentence patterns and voice |
| `check_plagiarism` | Check similarity and self-repetition |
| `get_detection_report` | Comprehensive detection report |
| `humanize_content` | Suggestions to make text more human |
| `batch_detect` | Detect multiple texts at once |
| `compare_versions` | Compare AI scores between versions |
| `get_ai_probability` | Quick AI probability score |

## AI Patterns Detected

🔴 **High Severity**: Elevated vocabulary, vague attribution, promotional tone, boilerplate intro, lack of specificity, buzzword density, no personal anecdotes, parallel negation

🟡 **Medium Severity**: Em dash overuse, rule of three, transition overuse, conclusion summary, hedge words, redundant phrasing, over-explanation, neutral stance, generic examples

🟢 **Low Severity**: Perfect grammar, list-heavy, template structure, symmetric structure, emotional flatness

## Integration with HumanizerMCP

This detector pairs perfectly with HumanizerMCP:

1. **Detect** → Identify AI patterns and their severity
2. **Humanize** → Apply fixes using HumanizerMCP's rule/LLM engine
3. **Compare** → Verify AI score reduction with `compare_versions`

## MCP Config

```json
{
  "mcpServers": {
    "ai-content-detector": {
      "command": "python",
      "args": ["server.py"],
      "cwd": "/path/to/ai-content-detector-mcp"
    }
  }
}
```


## 在线访问（推荐）

无需本地安装，直接在MCP客户端配置：

```json
{
  "mcpServers": {
    "ai-content-detector-mcp": {
      "url": "http://www.mzse.com/detector-mcp/"
    }
  }
}
```

## REST API

```bash
curl http://www.mzse.com/detector-mcp/
```

## 部署状态

| 项目 | 地址 |
|------|------|
| 域名 | http://www.mzse.com/detector-mcp/ |
| GitHub | https://github.com/zhaohongyuziranerran/ai-content-detector-mcp |

## License

MIT