agent-eligibility

Deterministic eligibility scoring for AI agents: CLEAR/HOLD/FAIL verdicts, published weights.

我该使用它吗

质量与安全性

A
描述质量
97%
模式完整度
62%
命名质量
100%
投毒风险
100%
权限匹配度
100%
协议合规性
100%

基于对工具定义和协议合规性的自动分析。

上下文开销

~828token 数(工具定义)
~1003 B典型响应大小
对注意力有中等影响(占 128k 上下文窗口的 0.65%)

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": {
    "agent-eligibility": {
      "url": "https://paynemuse.pythonanywhere.com/mcp"
    }
  }
}

远程端点

https://paynemuse.pythonanywhere.com/mcpstreamable-http

它能做什么

工具清单

工具(6)

🟢 只读🟡 写入🔴 删除⚪ 未知
⚪evaluate_eligibility(task, required_capabilities, agent)

Score one (agent, task) pair. Returns the scorer's full result dict as JSON text: decision (CLEAR/HOLD/FAIL), score 0-100, confidence, reasons, next_action, plus evaluation metadata. Unknown (null) evidence counts as unverified.

输入模式

{
  "type": "object",
  "properties": {
    "task": {
      "title": "Task",
      "type": "string"
    },
    "required_capabilities": {
      "items": {
        "type": "string"
      },
      "title": "Required Capabilities",
      "type": "array"
    },
    "agent": {
      "$ref": "#/$defs/AgentInput"
    }
  },
  "required": [
    "task",
    "required_capabilities",
    "agent"
  ],
  "$defs": {
    "AgentInput": {
      "properties": {
        "name": {
          "default": "unnamed-agent",
          "title": "Name",
          "type": "string"
        },
        "capabilities": {
          "items": {
            "type": "string"
          },
          "title": "Capabilities",
          "type": "array"
        },
        "evidence": {
          "$ref": "#/$defs/Evidence"
        }
      },
      "title": "AgentInput",
      "type": "object"
    },
    "Evidence": {
      "properties": {
        "payment_verified": {
          "anyOf": [
            {
              "type": "boolean"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Payment Verified"
        },
        "delivery_verified": {
          "anyOf": [
            {
              "type": "boolean"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Delivery Verified"
        },
        "policy_compatible": {
          "anyOf": [
            {
              "type": "boolean"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Policy Compatible"
        },
        "source_conflict": {
          "anyOf": [
            {
              "type": "boolean"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Source Conflict"
        },
        "certification_fresh": {
          "anyOf": [
            {
              "type": "boolean"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Certification Fresh"
        },
        "failure_rate": {
          "anyOf": [
            {
              "type": "number"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Failure Rate"
        },
        "historical_jobs": {
          "anyOf": [
            {
              "type": "integer"
            },
            {
              "type": "null"
            }
          ],
          "default": null,
          "title": "Historical Jobs"
        }
      },
      "title": "Evidence",
      "type": "object"
    }
  },
  "title": "evaluate_eligibilityArguments"
}
🟢get_scoring_table

Return the v1.0.0 deduction table, decision bands, confidence rules, and rules version as JSON, so a client can display the published rules.

输入模式

{
  "type": "object",
  "properties": {},
  "title": "get_scoring_tableArguments"
}
🟢list_paid_services

List the paid x402 services offered by Payne Machine Foundry (prices, endpoints, how to buy). The MCP tools themselves are free.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
🟡check_vat(country_code, vat_number)

Check an EU VAT number against the official VIES register. Returns valid (bool), the registered name and address, the VIES request identifier, source and checked_at. Mirrors the paid HTTP endpoint POST /api/check/vat.

输入模式

{
  "type": "object",
  "properties": {
    "country_code": {
      "type": "string"
    },
    "vat_number": {
      "type": "string"
    }
  },
  "required": [
    "country_code",
    "vat_number"
  ],
  "additionalProperties": false
}
🟡check_sanctions(name, type)

Screen a person or entity name against the OFAC SDN and EU consolidated (FSD) sanctions lists. Returns scored matches (name, list, program, type, score), the list versions used, and checked_at. Mirrors the paid HTTP endpoint POST /api/check/sanctions.

输入模式

{
  "type": "object",
  "properties": {
    "name": {
      "type": "string"
    },
    "type": {
      "type": "string",
      "enum": [
        "person",
        "entity"
      ]
    }
  },
  "required": [
    "name"
  ],
  "additionalProperties": false
}
🔴check_counterparty(name, country_code, vat_number)

Counterparty due-diligence verdict: sanctions screening plus a VAT registration check (when vat_number is given) composed into CLEAR / HOLD / FAIL with both component results. Mirrors the paid HTTP endpoint POST /api/check/counterparty.

输入模式

{
  "type": "object",
  "properties": {
    "name": {
      "type": "string"
    },
    "country_code": {
      "type": "string"
    },
    "vat_number": {
      "type": "string"
    }
  },
  "required": [
    "name"
  ],
  "additionalProperties": false
}

社区

评价此服务器

证据

最近观测

已验证未记录版本6 个工具