mcp-notifications
Enables sending real-time and webhook notifications when AI agents complete tasks, with support for Discord, Slack, Teams, Feishu, Ntfy, and custom webhooks.
README
MCP Notifications
This project is a spiritual successor to the original
mcp-server-notifierby tuberrabbit@gmail.com, and has been significantly rewritten and is now maintained by zudsniper.
A modern, scalable notification service that integrates with MCP (Model Context Protocol) to send real-time notifications and webhooks when AI agents complete tasks. Now featuring a monorepo architecture with SSE (Server-Sent Events) support, Next.js web interface, and PocketBase integration.

🚀 What's New in v3.0
- Monorepo Architecture: Organized as a workspace with separate apps and packages
- SSE Server: Real-time notifications via Server-Sent Events (Cloudflare Worker with Durable Objects)
- Web Dashboard: Next.js-based interface for monitoring and managing notifications
- PocketBase Integration: Persistent storage for webhooks, connections, and notification history
- Enhanced Provider Support: Improved webhook formatting for all major platforms
- Rate Limiting: Built-in rate limiting with KV storage
- Connection Management: Automatic cleanup and heartbeat for SSE connections
- TypeScript Throughout: Full TypeScript support with shared types package
Features
- Real-time Notifications: SSE-based push notifications for instant updates
- Webhook Notifications: Receive alerts when your AI agents complete tasks
- Multiple Webhook Providers: Support for Discord, Slack, Microsoft Teams, Feishu, Ntfy, and custom webhooks
- Image Support: Include images in notifications via Imgur
- Multi-Project Support: Efficiently manage notifications across different projects
- Easy Integration: Simple setup with AI tools like Cursor
- Customizable Messages: Send personalized notifications with title, body, and links
- Discord Embed Support: Send rich, customizable Discord embed notifications
- NTFY Template Support: Use pre-defined templates for status, questions, progress, and problems
- Notification History: Track all notifications with PocketBase storage
- Fallback Support: Configure backup webhooks for critical notifications
- Queue Processing: Reliable webhook delivery with queue-based processing
Architecture
The v3.0 architecture consists of multiple components working together:
- SSE Server (Cloudflare Worker): Manages real-time connections using Durable Objects
- Web Dashboard (Next.js): User interface for monitoring and configuration
- PocketBase: Persistent storage for webhooks, templates, and history
- Queue System: Reliable webhook delivery with retry logic
- Legacy MCP Server: Backward compatibility for existing integrations
Data Flow
User/AI Agent → SSE Server → Queue → Webhook Provider
↓ ↓
PocketBase Rate Limiter
↓
Web Dashboard
Project Structure
This is a monorepo organized with npm workspaces:
mcp-notifications/
├── apps/
│ ├── sse-server/ # Cloudflare Worker for SSE connections
│ └── web/ # Next.js dashboard
├── packages/
│ └── shared/ # Shared types and utilities
├── legacy/ # Legacy MCP server implementation
└── pocketbase/ # PocketBase schema and migrations
Installation
For Development
# Clone the repository
git clone https://github.com/zudsniper/mcp-notifications.git
cd mcp-notifications
# Install dependencies for all workspaces
npm install
# Build all packages
npm run build
Deploy SSE Server (Cloudflare Worker)
cd apps/sse-server
npm run deploy
Run Web Dashboard
cd apps/web
npm run dev
Legacy MCP Server
The original MCP server is still available in the legacy/ directory:
cd legacy
npm install
npm run build
Integration
SSE Client Integration
Connect to the SSE server from your application:
import { SSEClient } from '@mcp-notifications/shared';
const client = new SSEClient({
userId: 'user-123',
sseServerUrl: 'https://your-worker.workers.dev',
onNotification: (notification) => {
console.log('Received notification:', notification);
},
onConnectionStatus: (status) => {
console.log('Connection status:', status);
}
});
client.connect();
API Endpoints
Send notifications via the REST API:
curl -X POST https://your-worker.workers.dev/api/notify \
-H "Content-Type: application/json" \
-d '{
"userId": "user-123",
"message": {
"title": "Task Complete",
"body": "Your AI agent finished processing"
},
"webhookConfig": {
"type": "discord",
"url": "https://discord.com/api/webhooks/..."
}
}'
Legacy Cursor Integration
For the legacy MCP server:
{
"mcpServers": {
"notifier": {
"command": "node",
"args": [
"/path/to/legacy/build/index.js"
],
"env": {
"WEBHOOK_URL": "https://ntfy.sh/webhook-url-example",
"WEBHOOK_TYPE": "ntfy"
}
}
}
}
## Configuration
### SSE Server Configuration (Cloudflare Worker)
Configure in `apps/sse-server/wrangler.toml`:
```toml
name = "mcp-notifications-sse"
main = "src/index.ts"
compatibility_date = "2024-01-01"
[durable_objects]
bindings = [{ name = "SSE_CONNECTION_MANAGER", class_name = "SSEConnectionManager" }]
[[kv_namespaces]]
binding = "RATE_LIMIT_KV"
id = "your-kv-namespace-id"
[[queues.producers]]
binding = "WEBHOOK_QUEUE"
queue = "webhook-notifications"
[vars]
POCKETBASE_URL = "https://your-pocketbase-instance.com"
Web Dashboard Configuration
Configure in apps/web/.env.local:
NEXT_PUBLIC_SSE_SERVER_URL=https://your-worker.workers.dev
NEXT_PUBLIC_POCKETBASE_URL=https://your-pocketbase-instance.com
PocketBase Setup
- Deploy PocketBase instance
- Import schema:
pocketbase/pb_schema.json - Run migrations for initial templates
Supported Webhook Types
- Discord: Rich embeds with colors and fields
- Slack: Blocks and attachments
- Microsoft Teams: Adaptive cards
- Feishu: Interactive cards
- Ntfy: Priority levels and actions
- Generic JSON: Custom webhook format
Legacy Configuration
For the legacy MCP server, create legacy/webhook-config.json:
{
"webhook": {
"type": "discord",
"url": "https://discord.com/api/webhooks/your-webhook-url",
"name": "My Notifier"
},
"imgur": {
"clientId": "your-imgur-client-id"
}
}
See the Configuration Guide for full details and examples.
Usage
For detailed usage instructions, see the Usage Guide.
Available Tools
This package provides two tools for sending notifications:
notify- for simple notifications.full_notify- for more advanced notifications with all features.
notify
Send a simple notification with body, optional title, and optional template.
Input:
body: The main content of the notification message.title(optional): The title for the notification.template(optional): A predefined template to use (e.g., 'status', 'question', 'progress', 'problem').
Example:
// AI agent call
await run("notify", {
title: "Task Completed",
body: "I have finished the task."
});
full_notify
Send a detailed notification with advanced options like a link, image, priority, attachments, actions, and template data.
Input:
body: The main content of the notification message.title(optional): The title for the notification.link(optional): A URL to include in the notification.imageUrl(optional): The URL of an image to include.image(optional): The local file path for an image to upload to Imgur.priority(optional, ntfy only): Notification priority level from 1-5 (5 is the highest).attachments(optional, ntfy only): A list of URLs to attach to the notification.template(optional): A predefined template to use.templateData(optional): Data to be used with the selected template.actions(optional, ntfy only): Interactive action buttons for the notification.
Example:
// AI agent call
await run("full_notify", {
title: "Server Alert",
body: "Disk usage is high!",
priority: 5,
actions: [
{
action: "view",
label: "Open Grafana",
url: "https://grafana.example.com/d/abcdefg"
}
]
});
NTFY Templates
When using ntfy.sh as your webhook provider, you can use the following predefined templates with the template and templateData parameters:
status: For sending status updates.question: For asking questions.progress: For tracking the progress of long-running tasks.problem: For reporting errors or issues.
See docs/NOTIFICATIONS.md for more details on templates.
Development
Setting Up Development Environment
- Clone the repository:
git clone https://github.com/zudsniper/mcp-notifications.git
cd mcp-notifications
- Install dependencies for all workspaces:
npm install
- Build all packages:
npm run build
Development Commands
# Run SSE server locally
cd apps/sse-server
npm run dev
# Run web dashboard
cd apps/web
npm run dev
# Build shared package
cd packages/shared
npm run build
# Test legacy server
cd legacy
npm run test
Testing Webhooks
Test scripts are available in legacy/src/:
test-discord.js- Test Discord webhookstest-ntfy.js- Test ntfy notificationstest-slack.js- Test Slack webhooks
Deployment
Deploy SSE Server to Cloudflare
cd apps/sse-server
npm run deploy
Deploy Web Dashboard to Vercel
cd apps/web
vercel
Migration Guide
Migrating from v2.x to v3.0
- Legacy Support: The original MCP server is still available in
legacy/directory - New Features: Consider adopting SSE for real-time notifications
- Database: Set up PocketBase for persistent storage
- API Changes: Update API calls to use new endpoints
See MIGRATION.md for detailed migration instructions.
Performance
- SSE Connections: Support for thousands of concurrent connections via Durable Objects
- Rate Limiting: Built-in protection against abuse (10 requests/minute default)
- Queue Processing: Reliable webhook delivery with automatic retries
- Edge Computing: Global distribution via Cloudflare Workers
Security
- Authentication: Token-based auth for ntfy webhooks
- Rate Limiting: KV-based rate limiting per user
- CORS: Configurable CORS headers for web clients
- Input Validation: Zod schemas for all API inputs
License
MIT License - see LICENSE file for details.
Contributing
Contributions are welcome! Please feel free to submit a Pull Request.
Acknowledgments
- Original
mcp-server-notifierby tuberrabbit@gmail.com - Maintained and enhanced by zudsniper
- Built with Cloudflare Workers, Next.js, and PocketBase
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。