Loop Breaker

Catch AI agent loops: halt runs that repeat the same failing step or tool call.

我该使用它吗

质量与安全性

A
描述质量
100%
模式完整度
64%
命名质量
85%
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

~558token 数(工具定义)
~863 B典型响应大小
对注意力的影响极小(占 128k 上下文窗口的 0.44%)

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": {
    "loop": {
      "url": "https://loop.cloudforbots.com/mcp"
    }
  }
}

远程端点

https://loop.cloudforbots.com/mcpstreamable-http

它能做什么

工具清单

工具(4)

🟢 只读🟡 写入🔴 删除⚪ 未知
🟢check_step(run_id, tool, args, error, max_failures, ...)

Call AFTER each tool call or agent step with what you just did and whether it failed. Loop Breaker fingerprints the call (tool + normalized args + error) and returns verdict ok, loop_detected or halted. Pro: on a loop, halt=true; stop the run, do not retry, and tell your user what repeated. Free watch-only: never halts, returns would_halt and logs it for the weekly report.

输入模式

{
  "type": "object",
  "properties": {
    "run_id": {
      "type": "string",
      "description": "Any stable id for this agent run, e.g. \"nightly-sync-2026-10-15\". Use the same id for every step of the run (and its sub-agents)."
    },
    "tool": {
      "type": "string",
      "description": "The tool or step name you just ran, e.g. \"browser.click\" or \"git push\"."
    },
    "args": {
      "description": "The arguments you passed (object or string). Volatile values like timestamps and ids are normalized away."
    },
    "error": {
      "type": "string",
      "description": "The error message if the step failed. Omit or empty if it succeeded."
    },
    "max_failures": {
      "type": "integer",
      "description": "Same failing call this many times = loop. Default 3 (min 2, max 20)."
    },
    "max_repeats": {
      "type": "integer",
      "description": "Same call this many times in a row (even if it succeeds) = loop. Default 5 (min 3, max 50)."
    }
  },
  "required": [
    "run_id",
    "tool"
  ]
}
🟢loop_status(run_id)

Show a run's loops so far, whether it is halted, and its last 10 steps.

输入模式

{
  "type": "object",
  "properties": {
    "run_id": {
      "type": "string",
      "description": "Any stable id for this agent run, e.g. \"nightly-sync-2026-10-15\". Use the same id for every step of the run (and its sub-agents)."
    }
  },
  "required": [
    "run_id"
  ]
}
🟢weekly_report

Draft a plain-English report of the last 7 days for your key (or IP): runs, loops caught, wasted repeat steps, and the worst offenders. Read-only; sends nothing. Show it to your user.

输入模式

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

Show today's usage for your key or IP and the $49/month All Access upgrade link (cloudforbots.com/offer). Free tier: 500 calls/day. Not metered.

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4 个工具