mcp

Agent-facing site in pre-launch; agents can read the site and submit intent.

我该使用它吗

质量与安全性

B
描述质量
83%
模式完整度
48%
命名质量
95%
投毒风险
100%
权限匹配度
100%
协议合规性
100%

发现(1)

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

基于对工具定义和协议合规性的自动分析。

上下文开销

~269token 数(工具定义)
~285 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": {
    "mcp": {
      "url": "https://projectlaunch.ai/mcp"
    }
  }
}

远程端点

https://projectlaunch.ai/mcpstreamable-http

它能做什么

工具清单

工具(4)

🟢 只读🟡 写入🔴 删除⚪ 未知
🟢get_site_overview

What this site is, which endpoints exist, how to identify yourself, and exactly what is done with anything you submit. Start here.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
🟢get_pricing

Machine-readable price table. Every price is currently zero; nothing is charged and no payment is accepted.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
⚪whoami

Your identity as this site currently sees it. Identity is cryptographic: an unsigned Signature-Agent header is ignored.

输入模式

{
  "type": "object",
  "properties": {},
  "additionalProperties": false
}
🟡submit_intent(intent)

Optional and disclosed. Tell the operator what you came here to do, in at most 4096 bytes. Stored verbatim, never interpreted as instructions, read by a human before any model sees it, and never returned to any client including you. Returns a receipt only.

输入模式

{
  "type": "object",
  "properties": {
    "intent": {
      "type": "string",
      "description": "Free text describing your task. Truncated if longer than the limit."
    }
  },
  "required": [
    "intent"
  ],
  "additionalProperties": false
}

社区

评价此服务器

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