Skip to content
Projects

CCCpan/chinese-sensitive-words-mcp

View on GitHub

CCCpan/chinese-sensitive-words-mcp has 105 GitHub stars and 6 forks. GitHot measured 0 stars in 24 hours. Last checked 2026-11-16 19:47 UTC. Last push 2026-03-15.

How these counts are measured

Measured star growth

in 24h
0
in 7 days
Update delayed
in 15 days
Update delayed
in 1 month
Update delayed

Net star changes over overlapping periods, not daily totals. Missing measurements are left blank. Stars measure attention, not software quality.

About

中文敏感词/违禁词检测 MCP Server | 支持小红书/抖音/快手/B站 | 风险分级+安全替换建议 | 微信: chenganp | 邮箱: [email protected]

Developer Tools

README_EN.md

Original

Chinese Sensitive Words MCP Server

中文 | English

MCP Server for detecting sensitive/prohibited words in Chinese text. Built for social media content compliance on Xiaohongshu (Little Red Book), Douyin (TikTok China), Kuaishou, and Bilibili.

✨ Features

  • Massive dictionary, updated daily — Covers politics, pornography, violence, gambling, drugs, advertising law violations, medical claims, and more
  • Multi-platform support — Platform-specific dictionaries for Xiaohongshu, Douyin, Kuaishou, Bilibili
  • Risk level classification — High (account ban) / Medium (content throttling) / Low (recommended edit) / Tip
  • Replacement suggestions — Not just detection, also recommends safe alternative words
  • Homophone & variant detection — Catches "薇信→微信", "℡→电话" and similar evasion patterns
  • Skip-character detection — Catches "加 微 信", "最.好.的" with interference characters
  • Phone number & URL detection — Automatically flags contact info and external links
  • NER smart filtering — Reduces false positives for place names, person names, organizations

🛠️ Tools

Tool Description When to use
check_sensitive_words Detect sensitive/prohibited words in Chinese text Check marketing copy, product descriptions, live-streaming scripts for compliance
get_word_suggestions Get safe replacement suggestions for flagged words Fix detected violations with platform-safe alternatives

📦 Installation

Claude Desktop / Cursor / Windsurf

Add to your configuration file:

{
  "mcpServers": {
    "chinese-sensitive-words": {
      "command": "npx",
      "args": ["-y", "chinese-sensitive-words-mcp"]
    }
  }
}

Claude Code

claude mcp add chinese-sensitive-words -- npx -y chinese-sensitive-words-mcp

OpenClaw

npx mcporter config add chinese-sensitive-words-mcp --stdio "npx -y chinese-sensitive-words-mcp"

📖 Usage Examples

Check text for sensitive words

User: "Check this marketing copy for prohibited words"

AI calls check_sensitive_words and returns:

⚠️ Detected 3 sensitive word(s):

Risk summary: 🔴 High=0 | 🟡 Medium=2 | 🔵 Low=1

🟡 MEDIUM RISK (may cause content throttling)
- "最好用" — Category: Advertising extreme word → Suggested replacements: 很好用, 超好用
- "美白" — Category: Medical claim → Suggested replacements: 提亮, 焕亮

🔵 LOW RISK (recommend modification)
- "加微信" — Category: Traffic diversion → Suggested replacements: 私信咨询

Get replacement suggestions

User: "What can I use instead of 最好用?"

AI calls get_word_suggestions(keyword="最好用") and returns:

"最好用" (Extreme word replacement)
Suggested replacements: 很好用, 超好用, 非常好用, 特别好用

📋 Usage Limits

Tier Quota Notes
Free (no token) 100 requests/day Works out of the box
Registered (with token) Unlimited Set WORDSCHECK_ACCESS_TOKEN

🔧 Configuration

Basic (free, 100 requests/day)

{
  "mcpServers": {
    "chinese-sensitive-words": {
      "command": "npx",
      "args": ["-y", "chinese-sensitive-words-mcp"]
    }
  }
}

Registered user (unlimited)

{
  "mcpServers": {
    "chinese-sensitive-words": {
      "command": "npx",
      "args": ["-y", "chinese-sensitive-words-mcp"],
      "env": {
        "WORDSCHECK_ACCESS_TOKEN": "your_token_here"
      }
    }
  }
}

Custom service endpoint

{
  "mcpServers": {
    "chinese-sensitive-words": {
      "command": "npx",
      "args": ["-y", "chinese-sensitive-words-mcp"],
      "env": {
        "WORDSCHECK_API_BASE": "https://your-server.com/api",
        "WORDSCHECK_ACCESS_TOKEN": "your_token_here"
      }
    }
  }
}

📊 Risk Levels

Level Impact Examples
🔴 High Account ban / content removal Political, pornography, violence, phone numbers
🟡 Medium Content throttling / reduced reach Advertising law violations, medical claims
🔵 Low Recommended to modify Traffic diversion, promotional terms
💡 Tip Consider rewording Anxiety marketing, appearance-related terms

🤝 Contributing

Issues and Pull Requests are welcome! See CONTRIBUTING.md.

📄 License

MIT