Calendar MCP Server
A production-quality MCP server for calendar management with robust timezone handling, recurrence support, and idempotency guarantees.
README
MCP Calendar Server
A production-quality Model Context Protocol (MCP) server for calendar management with robust timezone handling, recurrence support, and idempotency guarantees.
🎯 What is MCP?
The Model Context Protocol is a standardized way for AI agents (like Claude, ChatGPT, etc.) to interact with external tools and data sources. This server exposes calendar operations as MCP tools, allowing AI assistants to:
- Create events with timezone-aware scheduling
- List events with automatic recurrence expansion
- Cancel individual instances or entire event series
✨ Features
🌍 Timezone Correctness (Top Priority)
Timezone handling is notoriously difficult. Here's how we handle it:
- Storage: All events stored in UTC (source of truth)
- Input: Accept ISO 8601 datetime + IANA timezone (e.g.,
"Asia/Kolkata") - Output: Return events in their original timezone
- DST Handling: Recurrence expansion happens in local time to preserve wall-clock semantics (e.g., "9 AM daily" stays at 9 AM even across DST transitions)
- Library: Uses Luxon for robust timezone math
Why UTC Storage?
- Single source of truth
- No ambiguity during DST transitions
- Easy comparison and sorting
- Portable across systems
Why Original Timezone Matters?
- Preserves user intent ("9 AM in Kolkata" not "3:30 AM UTC")
- Correct recurrence expansion (daily meetings stay at same wall time)
🔁 Recurrence Support
Supports repeating events with:
- Frequencies:
daily,weekly,monthly - Intervals: Every N days/weeks/months (e.g., "every 2 weeks")
- End Dates:
untilparameter (ISO 8601)
Implementation Design:
- Recurrence rules stored as JSON in the database
- Virtual instances generated on-demand during
calendar.list - No duplicate rows for recurring events
- Cancel individual instances via
cancellationstable - Safety limits: Max 1000 instances per query to prevent DOS
🔒 Idempotency
Prevents duplicate events from retries:
- Optional
idempotency_keyparameter incalendar.create - If key exists, returns existing event (no duplicate created)
- Unique constraint enforced at database level
💾 Storage
Uses SQLite with better-sqlite3:
- Synchronous API: Simpler, more reliable than async
- WAL Mode: Better concurrency
- Schema:
events: Core event data (title, start_utc, duration, timezone, recurrence)cancellations: Tracks cancelled instances of recurring events- Indexes on
start_utcandidempotency_keyfor performance
📦 Installation
# Clone or navigate to project directory
cd calender-mcp
# Install dependencies
npm install
# Build TypeScript
npm run build
🚀 Usage
Running the Server
npm start
The server runs on stdio (standard input/output) as per MCP specification.
Configuring with Claude Desktop
Add to your Claude Desktop config (claude_desktop_config.json):
{
"mcpServers": {
"calendar": {
"command": "node",
"args": ["C:/path/to/calender-mcp/dist/index.js"]
}
}
}
Example Tool Calls
1. Create a Simple Event
{
"tool": "calendar.create",
"arguments": {
"title": "Team Standup",
"start": "2026-01-20T09:00:00",
"end": "2026-01-20T09:30:00",
"tz": "Asia/Kolkata"
}
}
2. Create a Recurring Event
{
"tool": "calendar.create",
"arguments": {
"title": "Weekly Review",
"start": "2026-01-20T15:00:00",
"end": "2026-01-20T16:00:00",
"tz": "America/New_York",
"recurrence": {
"freq": "weekly",
"interval": 1,
"until": "2026-06-30T23:59:59"
},
"idempotency_key": "weekly-review-2026"
}
}
3. List Events
{
"tool": "calendar.list",
"arguments": {
"range_start": "2026-01-20T00:00:00Z",
"range_end": "2026-01-27T00:00:00Z"
}
}
4. Cancel an Entire Event Series
{
"tool": "calendar.cancel",
"arguments": {
"event_id": "abc-123-def-456"
}
}
5. Cancel a Specific Instance
{
"tool": "calendar.cancel",
"arguments": {
"event_id": "abc-123-def-456",
"instance_start_utc": "2026-01-27T10:00:00.000Z"
}
}
🏗️ Architecture
src/
├── index.ts # MCP server setup & tool registration
├── storage/
│ └── db.ts # SQLite schema & connection
├── time/
│ └── timezone.ts # UTC conversion utilities (Luxon)
├── recurrence/
│ └── expand.ts # Recurrence expansion logic
└── tools/
├── create.ts # calendar.create handler
├── list.ts # calendar.list handler
└── cancel.ts # calendar.cancel handler
🧪 Testing
Create a test script (test.mjs):
import { spawn } from 'child_process';
const server = spawn('node', ['dist/index.js']);
// Send MCP request
const request = {
jsonrpc: '2.0',
id: 1,
method: 'tools/call',
params: {
name: 'calendar.create',
arguments: {
title: 'Test Event',
start: '2026-01-20T10:00:00',
end: '2026-01-20T11:00:00',
tz: 'UTC'
}
}
};
server.stdin.write(JSON.stringify(request) + '\n');
server.stdout.on('data', (data) => {
console.log('Response:', data.toString());
});
Run: node test.mjs
🚧 Known Limitations
- No Sync: Events are stored locally only. No Google Calendar / Outlook sync.
- No Notifications: Server doesn't send reminders or alerts.
- Recurrence Complexity: Only supports simple rules (no "2nd Tuesday of month").
- Performance: Recurrence expansion is brute-force iteration (acceptable for <1000 instances).
- No Conflict Detection: Doesn't prevent overlapping events.
🔮 Future Extensions
- Calendar Sync: Integrate with Google Calendar API, Microsoft Graph
- Notifications: Email/SMS reminders via Twilio, SendGrid
- Search: Full-text search on event titles/notes
- Attachments: Store files/links with events
- Attendees: Multi-user support with invitations
- Conflict Detection: Warn about overlapping events
- Advanced Recurrence: RRULE support (RFC 5545)
📚 Design Decisions
Why Luxon over Moment/Date-fns?
- Immutable: Safer API, no accidental mutations
- IANA Timezone Support: Built-in, no extra plugins
- Modern: Active development, ESM-first
Why SQLite over JSON Files?
- ACID Guarantees: Atomic writes, no corruption
- Indexes: Fast lookups on
start_utc,idempotency_key - Constraints: Unique keys enforced at DB level
- Scalability: Handles thousands of events efficiently
Why Virtual Instances for Recurrence?
- Storage Efficiency: One row instead of hundreds
- Flexibility: Change recurrence rule retroactively
- Cancellations: Track exceptions cleanly
📄 License
ISC
Built with ❤️ for the MCP ecosystem
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。