ai-eval
Cloudflare Workers MCP server: ai-eval
我该使用它吗
质量与安全性
B
发现(2)
- LOW在 score_response 中
- LOW在 compare_responses 中
基于对工具定义和协议合规性的自动分析。
上下文开销
~214token 数(工具定义)
~567 B典型响应大小
对注意力的影响极小(占 128k 上下文窗口的 0.17%)
这是每次将服务器的工具加载到模型上下文窗口时所消耗的大致 token 数。数值越高,可用于其他任务的注意力就越少。
安装
一键安装
将以下内容添加到你的 `claude_desktop_config.json` 文件中:
{
"mcpServers": {
"ai-eval": {
"url": "https://api.lazy-mac.com/ai-eval/mcp"
}
}
}远程端点
https://api.lazy-mac.com/ai-eval/mcpstreamable-http它能做什么
工具清单
工具(3)
🟢 只读🟡 写入🔴 删除⚪ 未知
⚪score_response(prompt, response, criteria)
Score an AI response against a prompt using heuristic metrics (length, relevance, structure, completeness)
输入模式
{
"type": "object",
"properties": {
"prompt": {
"type": "string",
"description": "The original prompt/question"
},
"response": {
"type": "string",
"description": "The AI response to evaluate"
},
"criteria": {
"type": "array",
"items": {
"type": "string"
},
"description": "Optional keywords that should appear in response"
}
},
"required": [
"prompt",
"response"
]
}⚪compare_responses(prompt, responses)
Compare and rank multiple AI responses to the same prompt
输入模式
{
"type": "object",
"properties": {
"prompt": {
"type": "string"
},
"responses": {
"type": "array",
"minItems": 2,
"items": {
"type": "string"
}
}
},
"required": [
"prompt",
"responses"
]
}🟢text_metrics(text)
Get text quality metrics: word count, sentence count, estimated tokens, readability grade
输入模式
{
"type": "object",
"properties": {
"text": {
"type": "string"
}
},
"required": [
"text"
]
}社区
证据
最近观测
已验证未记录版本3 个工具
已验证未记录版本3 个工具