cassandra-predict

Cassandra prediction hub: one MCP door over govcon forecasts (recompete, contractor signals).

我该使用它吗

质量与安全性

A
描述质量
100%
模式完整度
65%
命名质量
90%
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

~264token 数(工具定义)
~446 B典型响应大小
对注意力的影响极小(占 128k 上下文窗口的 0.21%)

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": {
    "cassandra-predict": {
      "url": "https://predict.dropwatchhq.com/mcp"
    }
  }
}

远程端点

https://predict.dropwatchhq.com/mcpstreamable-http

它能做什么

工具清单

工具(2)

🟢 只读🟡 写入🔴 删除⚪ 未知
⚪predict(domain, entity)

Cassandra unified predictor. From leading signals in public data, returns what is about to happen to an entity — weeks early — across four verticals. domain selects the vertical: 'recall' (consumer product / vehicle / drug / device recall risk from complaint velocity), 'bio' (biotech NIH→trial→SEC catalyst chain), 'recompete' (federal contract recompete-winner shift), 'ma' (government contractor acquisition target). entity is the name to look up (e.g. 'Tesla Model Y', 'Moderna', a contractor or agency name). Returns a leading-signal prediction with lead time — NOT a guarantee, and NOT a claim of any official action.

输入模式

{
  "type": "object",
  "properties": {
    "domain": {
      "type": "string",
      "enum": [
        "recall",
        "bio",
        "recompete",
        "ma"
      ],
      "description": "Prediction vertical to route to."
    },
    "entity": {
      "type": "string",
      "description": "Entity name / ticker / id to predict on."
    }
  },
  "required": [
    "domain",
    "entity"
  ]
}
🟢list_domains

List Cassandra prediction verticals available via predict(domain, entity), with what each predicts and its data sources.

输入模式

{
  "type": "object",
  "properties": {}
}

社区

评价此服务器

证据

最近观测

已验证未记录版本2 个工具
已验证未记录版本2 个工具
已验证未记录版本2 个工具