NumProof
Deterministic signed verification of numeric & financial claims for AI agents & spreadsheets.
我该使用它吗
质量与安全性
B
发现(1)
- LOW在 diff_rows 中
基于对工具定义和协议合规性的自动分析。
上下文开销
~251token 数(工具定义)
~314 B典型响应大小
对注意力的影响极小(占 128k 上下文窗口的 0.20%)
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": {
"numproof": {
"command": "uvx",
"args": [
"numproof"
]
}
}
}可运行的软件包
pypinumproof
0.1.1stdio远程端点
https://numproof.com/mcpstreamable-http它能做什么
工具清单
工具(4)
🟢 只读🟡 写入🔴 删除⚪ 未知
⚪verify_claim(claim)
Exactly verify a math/finance claim (VERIFY/REFUTE/ABSTAIN) with a counterexample when false. Use before trusting any AI-produced number, sum, percentage, or formula.
输入模式
{
"type": "object",
"properties": {
"claim": {
"type": "string"
}
},
"required": [
"claim"
]
}⚪audit_rows(rows)
Audit spreadsheet-like rows for footing, balance-sheet ties, common margins, and cell provenance.
输入模式
{
"type": "object",
"properties": {
"rows": {
"type": "array"
}
},
"required": [
"rows"
]
}⚪diff_rows(rows_before, rows_after)
Compare two report versions by numeric row labels with provenance.
输入模式
{
"type": "object",
"properties": {
"rows_before": {
"type": "array"
},
"rows_after": {
"type": "array"
}
},
"required": [
"rows_before",
"rows_after"
]
}⚪covenant_rules(rows, rules, rule_pack)
Evaluate threshold/covenant rules over spreadsheet-like rows with provenance. Use either rules or rule_pack.
输入模式
{
"type": "object",
"properties": {
"rows": {
"type": "array"
},
"rules": {
"type": "array"
},
"rule_pack": {
"type": "string"
}
},
"required": [
"rows"
]
}社区
证据
最近观测
已验证未记录版本4 个工具
已验证未记录版本4 个工具
已验证未记录版本4 个工具