MCP MeasureMCP Measure

索引已觀測

llm-output-quality-monitor

Cloudflare Workers MCP server: llm-output-quality-monitor

我該用這個嗎

品質與安全性

A
說明品質
94%
結構描述完整度
100%
命名品質
80%
汙染風險
100%
權限相符程度
100%
協定合規性
100%

發現項目（1）

  • LOWTool 'hallucination_scorer' description lacks action verb在 hallucination_scorer 中

根據工具定義與協定合規性的自動化分析。

上下文成本

~421Token（工具定義）
~756 B典型回應大小
極小的注意力影響（128k 上下文的 0.33%）

這是每次將伺服器的工具載入模型上下文時所消耗的約略 token 數量。數量越高，可用於其他工作的注意力就越少。

安裝

一鍵安裝

將以下內容加入你的 `claude_desktop_config.json` 檔案：

{
  "mcpServers": {
    "llm-output-quality-monitor": {
      "url": "https://api.lazy-mac.com/llm-output-quality-monitor/mcp"
    }
  }
}

遠端端點

https://api.lazy-mac.com/llm-output-quality-monitor/mcpstreamable-http

它能做什麼

工具清單

工具（5）

🟢 唯讀🟡 寫入🔴 刪除⚪ 未知
⚪quality_validator(response, minLength, maxLength, strictFormat)

Validate LLM response quality based on length, format, and structure

輸入結構描述

{
  "type": "object",
  "properties": {
    "response": {
      "type": "string",
      "description": "LLM response to validate"
    },
    "minLength": {
      "type": "number",
      "description": "Minimum response length (default: 10)"
    },
    "maxLength": {
      "type": "number",
      "description": "Maximum response length (default: 10000)"
    },
    "strictFormat": {
      "type": "boolean",
      "description": "Enforce punctuation and capitalization"
    }
  },
  "required": [
    "response"
  ]
}
⚪drift_detector(currentResponse, previousResponse, threshold)

Detect quality drift between current and previous LLM responses

輸入結構描述

{
  "type": "object",
  "properties": {
    "currentResponse": {
      "type": "string",
      "description": "Current LLM response"
    },
    "previousResponse": {
      "type": "string",
      "description": "Previous LLM response"
    },
    "threshold": {
      "type": "number",
      "description": "Drift threshold (0-1, default: 0.15)"
    }
  },
  "required": [
    "currentResponse",
    "previousResponse"
  ]
}
⚪hallucination_scorer(response, context)

Pattern-based heuristic risk scoring for LLM responses (0-100). Detects linguistic signals such as contradictory assertions, unsourced claims, and uncertainty markers. Not a semantic hallucination detector.

輸入結構描述

{
  "type": "object",
  "properties": {
    "response": {
      "type": "string",
      "description": "LLM response to analyze"
    },
    "context": {
      "type": "string",
      "description": "Reference context for grounding"
    }
  },
  "required": [
    "response"
  ]
}
⚪schema_enforcer(response, schema)

Validate JSON response against schema

輸入結構描述

{
  "type": "object",
  "properties": {
    "response": {
      "type": "string",
      "description": "JSON response to validate"
    },
    "schema": {
      "type": "object",
      "description": "JSON schema definition"
    }
  },
  "required": [
    "response",
    "schema"
  ]
}
🟢consistency_check(responses)

Check consistency across multiple LLM responses

輸入結構描述

{
  "type": "object",
  "properties": {
    "responses": {
      "type": "array",
      "items": {
        "type": "string"
      },
      "description": "Array of responses to compare"
    }
  },
  "required": [
    "responses"
  ]
}

社群

為此伺服器評分

證據

近期觀測

已驗證未記錄版本5 個工具
已驗證未記錄版本5 個工具
已驗證未記錄版本5 個工具