Skip to main content
Glama
deenesjakoruzh
nicholasemccormick

LoopIn MCP Server

by nicholasemccormick
OverviewSchemaRelated ServersScoreDiscussions
TypeScript
Remote

LoopIn

用于 AI 智能体的人机交互中断 API。

LoopIn 让 AI 智能体能够在关键决策点暂停，请求人工审核，并根据人类的决定恢复执行。不再让智能体独自做出高风险决策。

为什么选择 LoopIn？

自主智能体功能强大——直到它们遇到不应在没有人类参与的情况下做出的决策。LoopIn 为每个智能体提供了一个安全出口：

  1. 智能体到达决策点 → 调用 POST /interrupts

  2. LoopIn 通知人类 → 通过 Webhook 或直接审核 URL

  3. 人类批准或拒绝 → 通过审核页面或 POST /interrupts/:id/decide

  4. 智能体轮询决策 → GET /interrupts/:id

  5. 智能体恢复 → 使用该决策（以及人类提供的任何修改后的参数）

Related MCP server: final-approval

端点

POST /interrupts

创建新的中断请求。

请求：

{
  "agentId": "agent-payments-v2",
  "userId": "user-123",
  "action": "Transfer $4,200 to vendor account ending in 9821",
  "context": {
    "vendor": "Acme Supplies",
    "amount": 4200,
    "currency": "USD",
    "invoiceId": "INV-2024-0892",
    "accountLast4": "9821"
  },
  "urgency": "high",
  "expiresIn": 1800,
  "callbackUrl": "https://my-agent.example.com/webhooks/loopin"
}

响应：

{
  "interruptId": "3f4e2a1b-...",
  "status": "pending",
  "expiresAt": "2024-04-12T14:30:00Z",
  "reviewUrl": "http://localhost:3002/review/3f4e2a1b-..."
}

GET /interrupts/:interruptId

轮询当前状态和决策。

响应（待处理）：

{
  "interruptId": "3f4e2a1b-...",
  "status": "pending",
  "action": "Transfer $4,200 to vendor account ending in 9821",
  "context": { ... },
  "urgency": "high",
  "createdAt": "2024-04-12T13:00:00Z",
  "expiresAt": "2024-04-12T14:30:00Z",
  "reviewUrl": "http://localhost:3002/review/3f4e2a1b-..."
}

响应（已解决）：

{
  "interruptId": "3f4e2a1b-...",
  "status": "approved",
  "decision": "approved",
  "decidedAt": "2024-04-12T13:07:22Z",
  "reason": "Invoice verified, proceed."
}

POST /interrupts/:interruptId/decide

提交人类决策。

请求：

{
  "decision": "approved",
  "reason": "Invoice verified, proceed.",
  "modifiedParams": { "amount": 4200 }
}

响应：

{
  "interruptId": "3f4e2a1b-...",
  "status": "resolved",
  "decision": "approved",
  "decidedAt": "2024-04-12T13:07:22Z"
}

GET /interrupts/pending/:userId

列出所有等待用户审核的待处理中断。

响应：

[
  {
    "interruptId": "3f4e2a1b-...",
    "action": "Transfer $4,200 to vendor...",
    "urgency": "high",
    "createdAt": "...",
    "expiresAt": "...",
    "reviewUrl": "..."
  }
]

DELETE /interrupts/:interruptId

取消待处理的中断（智能体不再需要该决策）。

GET /analytics/:userId

使用统计信息。

响应：

{
  "userId": "user-123",
  "totalInterrupts": 47,
  "approvalRate": 0.83,
  "avgResponseTimeMs": 142000,
  "byStatus": { "approved": 39, "rejected": 8 },
  "byUrgency": { "high": 12, "medium": 28, "low": 7 },
  "topActionTypes": [
    { "action": "Send payment", "count": 18 },
    { "action": "Delete records", "count": 9 }
  ]
}

审核页面

每个中断都有一个人类可读的审核 URL：

GET /review/:interruptId

这将渲染一个 HTML 页面，显示：

  • 智能体想要做什么

  • 所有上下文数据（格式化的 JSON）

  • 紧急程度徽章

  • 批准 / 拒绝 按钮

  • 可选的理由文本框

将此 URL 分享给需要审核请求的任何人。默认无需登录。

MCP 工具

LoopIn MCP 服务器公开了 6 个工具，供任何兼容 MCP 的 AI 客户端（Claude Desktop、Cursor 等）使用：

工具

描述

create_interrupt

智能体创建新的中断请求

get_interrupt_status

智能体轮询决策

list_pending_interrupts

人类查看需要审核的内容

decide_interrupt

人类批准或拒绝

cancel_interrupt

智能体取消待处理请求

get_interrupt_analytics

使用统计信息

MCP 服务器设置 (stdio)

{
  "mcpServers": {
    "loopin": {
      "command": "npx",
      "args": ["-y", "@colossal-api/loopin-mcp"],
      "env": {
        "LOOPIN_API_URL": "https://your-loopin-instance.railway.app",
        "LOOPIN_API_KEY": "your-key"
      }
    }
  }
}

智能体使用模式

1. Agent reaches a decision point
   → POST /interrupts { agentId, userId, action, context, urgency }
   ← { interruptId, reviewUrl }

2. Agent saves interruptId and pauses
   → (optionally: notify human via other channels with reviewUrl)

3. Agent polls until resolved
   → GET /interrupts/:interruptId
   ← { status: "pending" }   ← keep polling
   ← { status: "approved", decision, modifiedParams }  ← resume

4. Agent resumes execution
   → use modifiedParams if provided, otherwise proceed as planned

人类使用模式

选项 A — 审核 URL（最简单）

  1. 从智能体接收 reviewUrl（通过电子邮件、Slack 等）

  2. 在任何浏览器中打开该 URL

  3. 查看上下文，点击“批准”或“拒绝”，添加可选理由

  4. 完成 — 智能体将在下一次轮询时获得决策

选项 B — 列出待处理（仪表板）

  1. GET /interrupts/pending/:userId — 查看所有打开的请求

  2. 打开单独的 reviewUrl 或直接调用 POST /interrupts/:id/decide

环境变量

变量

默认值

描述

PORT

3002

API 服务器端口

LOOPIN_BASE_URL

http://localhost:3002

审核链接的公共基础 URL

API_KEY_SECRET

(无)

可选：要求所有请求包含 X-API-Key 标头

Colossal API 产品组合

LoopIn 是 Colossal API 基础设施 API 套件的一部分，专为 AI 智能体设计：

  • SubRadar — 订阅检测与取消

  • MeetSync — 日程协商与安排

  • LoopIn — 人机交互中断与审批

所有产品共享相同的设计理念：简单的 REST API 和 MCP 服务器封装，以便智能体可以原生使用它们。

This server cannot be installed

A
license - permissive license
-
quality - not tested
D
maintenance

How are these scores calculated?

Maintenance

–Maintainers
–Response time
–Release cycle
–Releases (12mo)
Commit activity

Resources

Unclaimed servers have limited discoverability.

Looking for Admin?

If you are the server author, to access and configure the admin panel.

Related MCP Servers

  • vantagate-mcp-server

    Aderix
    A
    license
    -
    quality
    D
    maintenance
    Human-in-the-Loop authorization gateway for AI Agents. Securely pause MCP workflows and route high-risk actions to human approvers via Slack or Email.
    117
    1
    MIT

  • final-approval

    finalapproval
    A
    license
    A
    quality
    D
    maintenance
    Human-in-the-loop approval gate for AI agents. Your agent calls submit_approval before any irreversible action; a human reviews on a branded page; a signed webhook fires back with the decision.
    1
    1
    MIT

  • Approval Gate

    Keldrik
    A
    license
    -
    quality
    C
    maintenance
    Provides a human approval gate for AI agents, enabling interactive inline cards for approving, editing, or rejecting actions before they are executed.
    MIT

View all related MCP servers

Related MCP Connectors

  • MeatSpace

    Human-in-the-loop for AI agents. Submit choices, get a human decision.

  • Lassare

    Human-in-the-loop for AI coding agents — ask questions, get approvals via Slack.

  • NeedHuman

    Human-as-a-Service for AI agents. Delegate tasks that need a real human, get results via API.

View all MCP Connectors

Latest Blog Posts

MCP directory API

We provide all the information about MCP servers via our MCP API.

curl -X GET 'https://glama.ai/api/mcp/v1/servers/nicholasemccormick/loopin-mcp'

If you have feedback or need assistance with the MCP directory API, please join our Discord server