HAR Analyzer MCP Server

┌──────────────┐         ┌────────────────────┐         ┌──────────────┐
│  Chrome /    │  导出    │   HAR MCP Server   │  结构化  │   Trae AI    │
│  DevTools    │ ──────> │  (本地：清洗+索引)  │ ──────> │  (只读摘要)  │
└──────────────┘  .har   └────────────────────┘  JSON   └──────────────┘

har_mcp/
├── server.py           # MCP Server 主入口，注册 8 个 Tool
├── har_parser.py       # HAR 流式解析 + 数据清洗模块
├── har_store.py        # 内存 / SQLite 索引存储模块（线程安全）
├── filters.py          # 过滤规则（静态资源剔除、API 识别、Header 清洗、Body 处理、标签生成）
├── summarizer.py       # 摘要生成器（把 Entry 压缩为 AI 友好格式 + 加密/签名模式检测）
├── sanitizer.py        # 字符串脱敏模块（支持多种脱敏源）
├── config.py           # 全局配置常量（截断阈值、白名单、正则模式）
├── Sensitive-lexicon/  # git clone 下来的敏感词库（.gitignore 已忽略）
│   └── sensitive-lexicon.txt
├── wordlist.txt        # 自定义敏感词库（用户维护，不进版本控制）
├── requirements.txt    # Python 依赖
├── mcp.json            # Trae MCP 配置（python 直接运行）
├── mcp_uv.json         # Trae MCP 配置（uv 管理环境）
├── .gitignore          # Git 忽略规则
├── LICENSE             # 许可证（建议补充）
└── README.md           # 本文件

cd har_mcp
pip install -r requirements.txt

cd har_mcp
uv sync
# 或
uv pip install -r requirements.txt

conda create -n har-mcp python=3.11
conda activate har-mcp
pip install -r requirements.txt

python -c "from har_store import HarStore; store = HarStore(); print('✅ HarStore 初始化成功')"

cd har_mcp
git clone https://github.com/konsheng/Sensitive-lexicon.git

har_mcp/
├── server.py
├── sanitizer.py
├── Sensitive-lexicon/
│   └── sensitive-lexicon.txt   ← sanitizer 默认读取
└── wordlist.txt                ← 你的额外自定义词（可选）

export HAR_MCP_LEXICON=/path/to/Sensitive-lexicon/sensitive-lexicon.txt   # macOS/Linux
$env:HAR_MCP_LEXICON = "D:\Sensitive-lexicon\sensitive-lexicon.txt"        # PowerShell

{
  "mcpServers": {
    "har-analyzer": {
      "command": "python",
      "args": ["/absolute/path/to/har_mcp/server.py"],
      "env": {
        "PYTHONPATH": "/absolute/path/to/har_mcp"
      }
    }
  }
}

{
  "mcpServers": {
    "har-analyzer": {
      "command": "uv",
      "args": ["run", "--project", "/absolute/path/to/har_mcp", "python", "server.py"],
      "env": {}
    }
  }
}

{
  "mcpServers": {
    "har-analyzer": {
      "command": "py",
      "args": ["-3.11", "E:/Har_MCP/server.py"],
      "env": {
        "PYTHONPATH": "E:/Har_MCP"
      }
    }
  }
}

{
  "mcpServers": {
    "har-analyzer": {
      "command": "C:/Python311/python.exe",
      "args": ["E:/Har_MCP/server.py"],
      "env": {
        "PYTHONPATH": "E:/Har_MCP"
      }
    }
  }
}

  ┌─────────────────────────────────────────┐
  │ 1. har_load(filepath="...har")          │ ← 必须先调
  │    → 获得 total_entries / api_count     │
  └──────────────┬──────────────────────────┘
                 │
                 ▼
  ┌─────────────────────────────────────────┐
  │ 2. har_summary()                        │ ← 获得全局"地图"
  │    → endpoint_groups[]                  │
  │    → 找出可疑端点（status>=400, signed）│
  └──────────────┬──────────────────────────┘
                 │
        ┌────────┴────────┐
        ▼                 ▼
  ┌──────────┐      ┌─────────────────┐
  │ 3a.      │      │ 3b.             │
  │ har_     │      │ har_            │
  │ detect_  │      │ extract_        │
  │ patterns │      │ auth            │
  │ 加密特征 │      │ 认证机制        │
  └────┬─────┘      └────────┬────────┘
       │                     │
       └──────────┬──────────┘
                  ▼
  ┌─────────────────────────────────────────┐
  │ 4. har_search(...)                      │ ← 按 URL/方法/标签精确定位
  │    → 拿到 entry_id 列表                 │
  └──────────────┬──────────────────────────┘
                 │
                 ▼
  ┌─────────────────────────────────────────┐
  │ 5. har_get_entry(id)                    │ ← 按需查看详情
  │    或 har_get_entries_batch(ids)        │
  └─────────────────────────────────────────┘

{
  "status": "ok",
  "total_entries": 1247,
  "indexed_entries": 89,
  "skipped_static": 1158,
  "api_count": 89,
  "unique_hosts": ["api.example.com"],
  "time_range": {"start": "2025-01-15T10:00:00Z", "end": "2025-01-15T10:30:00Z"},
  "load_time_seconds": 2.34,
  "hint": "HAR 已加载完毕。建议先调用 har_summary ..."
}

{
  "method": "POST",
  "url_pattern": "https://api.example.com/api/v2/users/{id}",
  "count": 12,
  "status_codes": [200, 401, 500],
  "req_content_type": "application/json",
  "res_content_type": "application/json",
  "avg_time_ms": 234.5,
  "has_auth_header": true,
  "req_body_structure": {"username": "string", "password": "string"},
  "res_body_structure": {"code": "integer", "data": {"token": "string", "expire": "integer"}},
  "entry_ids": [42, 87, 156, ...]
}

{
  "auth_headers_found": {
    "authorization": {
      "sample_schemes": ["Bearer"],
      "entry_ids": [42, 87],
      "token_length_samples": [256, 248]
    },
    "x-sign": {
      "sample_schemes": ["none"],
      "entry_ids": [42, 87, 156],
      "token_length_samples": [64, 64, 64]
    }
  },
  "cookies_found": {
    "session_id": {"entry_count": 89}
  },
  "patterns_note": "authorization 使用 Bearer 方案，token 长度在 248-256 之间变化；x-sign 长度稳定在 64 字符，疑似 HMAC-SHA256 签名。"
}

# har_parser.py 核心
with open(filepath, 'rb') as f:
    entries = ijson.items(f, 'log.entries.item')
    for entry in entries:
        yield _clean_entry(entry)

export HAR_MCP_LEXICON=/path/to/Sensitive-lexicon/sensitive-lexicon.txt   # macOS/Linux
$env:HAR_MCP_LEXICON = "D:\Sensitive-lexicon\sensitive-lexicon.txt"        # PowerShell

# 主敏感词库（Sensitive-lexicon 的 sensitive-lexicon.txt）
export HAR_MCP_LEXICON=/path/to/Sensitive-lexicon/sensitive-lexicon.txt   # macOS/Linux
$env:HAR_MCP_LEXICON = "D:\Sensitive-lexicon\sensitive-lexicon.txt"        # PowerShell

# 额外自定义词库（可选，每行一词）
export HAR_MCP_WORDLIST=/path/to/your/wordlist.txt   # macOS/Linux
$env:HAR_MCP_WORDLIST = "D:\my-wordlist.txt"          # PowerShell

cd har_mcp
python -c "from har_store import HarStore; store = HarStore(); print('OK')"

import asyncio
from server import call_tool

async def test():
    result = await call_tool("har_load", {"filepath": "/path/to/test.har"})
    print(result[0].text)  # TextContent

asyncio.run(test())

from sanitizer import sanitize_string, sanitize_dict, get_sanitizer_status

print(get_sanitizer_status())
text, count = sanitize_string("这是一段含敏感词的文本")
print(f"命中词数: {count}, 结果: {text}")

# 测试嵌套 JSON
obj, count = sanitize_dict({"msg": "一段需要脱敏的文本", "code": 200})
print(f"命中词数: {count}, 结果: {obj}")