dav-mcp
Transform AI agents into orchestrating assistants managing calendars, contacts, and tasks via CalDAV, CardDAV, and VTODO protocols.
README
dav-mcp
Give your AI agents the power of organization — Transform them into orchestrating assistants managing calendars, contacts, and tasks.
Built on 26 production-ready tools spanning CalDAV, CardDAV, and VTODO protocols.
Quick Start
Claude Desktop / Cursor (Local)
Add to your MCP config file:
{
"mcpServers": {
"dav-mcp": {
"command": "npx",
"args": ["-y", "dav-mcp"],
"env": {
"CALDAV_SERVER_URL": "https://dav.example.com",
"CALDAV_USERNAME": "your_username",
"CALDAV_PASSWORD": "your_password"
}
}
}
}
Config file locations:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
Restart Claude Desktop after adding the configuration.
n8n (Remote HTTP)
Start the HTTP server:
CALDAV_SERVER_URL=https://dav.example.com \
CALDAV_USERNAME=your_username \
CALDAV_PASSWORD=your_password \
BEARER_TOKEN=your-secret-token \
npx dav-mcp --http
Then in n8n:
- Add AI Agent node
- Add MCP Client Tool node and connect to AI Agent
- Configure:
- MCP Endpoint:
http://localhost:3000/mcp - Authentication: Bearer
- Token: your-secret-token
- MCP Endpoint:
Custom port:
npx dav-mcp --http --port=8080
Docker
git clone https://github.com/PhilflowIO/dav-mcp.git
cd dav-mcp
cp .env.example .env
# Edit .env with your credentials
docker-compose up
The Orchestration
When partial tools force your AI to improvise, complete tools let it execute precise operations across all components.
| Capability | dav-mcp | Most MCPs |
|---|---|---|
| Calendar Management | Full CRUD (11 tools) | Create + list only (2-3 tools) |
| Contact Management | Complete CardDAV (8 tools) | Often missing entirely |
| Task Management | Full VTODO support (7 tools) | Rarely included |
| Field-Based Updates | All RFC properties + custom fields | Rarely available |
| Server-Side Filtering | Efficient queries | Dumps all data |
| Multi-Provider | Any CalDAV/CardDAV server | Limited provider support |
| Total Tools | 26 tools | 2-6 tools |
Available Tools (26 Total)
CalDAV Tools (11 tools)
- list_calendars - List all available calendars
- list_events - List ALL events (use calendar_query for filtered searches)
- create_event - Create a new calendar event
- update_event - PREFERRED: Update any event field (SUMMARY, LOCATION, DTSTART, STATUS, custom X-* properties)
- update_event_raw - Update event with raw iCal data (advanced)
- delete_event - Delete an event permanently
- calendar_query - PREFERRED: Search and filter events efficiently by text, date range, or location
- make_calendar - Create a new calendar collection
- update_calendar - Update calendar properties (display name, description, color, timezone)
- delete_calendar - Permanently delete a calendar and all its events
- calendar_multi_get - Batch fetch multiple specific events by URLs
CardDAV Tools (8 tools)
- list_addressbooks - List all available address books
- list_contacts - List ALL contacts (use addressbook_query for filtered searches)
- create_contact - Create a new contact (vCard)
- update_contact - PREFERRED: Update any contact field (FN, EMAIL, TEL, ORG, ADR, custom X-* properties)
- update_contact_raw - Update contact with raw vCard data (advanced)
- delete_contact - Delete a contact permanently
- addressbook_query - PREFERRED: Search and filter contacts efficiently by name, email, or organization
- addressbook_multi_get - Batch fetch multiple specific contacts by URLs
VTODO Tools (7 tools)
- list_todos - List ALL todos/tasks (use todo_query for filtered searches)
- create_todo - Create a new todo/task with optional due date, priority, status
- update_todo - PREFERRED: Update any todo field (SUMMARY, STATUS, PRIORITY, DUE, PERCENT-COMPLETE, custom X-* properties)
- update_todo_raw - Update todo with raw VTODO iCal data (advanced)
- delete_todo - Delete a todo/task permanently
- todo_query - PREFERRED: Search and filter todos efficiently by status/due date
- todo_multi_get - Batch fetch multiple specific todos by URLs
Real-World Applications
n8n Automation Workflows
- Meeting Management: "Show me all Friday meetings" → calendar_query with date filter returns only relevant events
- Contact Search: "Find everyone at Google" → addressbook_query with org filter finds matches efficiently
- Task Reporting: "Show overdue high-priority tasks" → todo_query with filters returns specific results
- Scheduled Cleanup: Daily cron job deletes completed tasks using targeted queries
Claude Desktop Integration
- Quick Event Creation: "Create team meeting tomorrow 2 PM" → create_event executes immediately
- Contact Lookup: "What's Sarah's email?" → addressbook_query with name filter finds contact
- Calendar Overview: "What's on my calendar next week?" → calendar_query with date range shows events
- Calendar Management: "Create a new calendar called Project Luna" → make_calendar creates collection
Works Across All Major Providers
Works with any CalDAV/CardDAV server that follows RFC 4791 and RFC 6352:
- Nextcloud - Full support
- Baikal - Full support
- Radicale - Full support
- iCloud - Works with app-specific password
- Any RFC-compliant server - Standard protocol support
CI/CD and Dependabot credentials
The GitHub Actions workflows expect the following secrets/variables:
| Name | Type | Purpose |
|---|---|---|
GHCR_USERNAME |
Secret (optional) | Username for GHCR login. Defaults to ${{ github.actor }} if not set. |
GH_PACKAGES_TOKEN |
Secret (required) | PAT with write:packages + read:packages for the classic GitHub Packages Docker registry. |
GH_PACKAGES_USERNAME |
Secret (optional) | Username for GitHub Packages login. Defaults to ${{ github.actor }} if not set. |
GHCR_REGISTRY |
Repository variable (optional) | Override registry host for GHCR (defaults to ghcr.io). |
GITHUB_PACKAGES_REGISTRY |
Repository variable (optional) | Override registry host for the legacy GitHub Packages Docker registry (defaults to docker.pkg.github.com). |
GH_PAT_TOKEN |
Secret (required for Dependabot) | PAT with read:packages to authenticate to the GitHub npm registry (npm.pkg.github.com). |
- Node.js version for CI workflows is read from the
engines.nodefield inpackage.jsonso Dependabot-managed updates stay in sync. - Dependabot is configured to update both direct and indirect npm dependencies in
package.json/package-lock.jsonthrough the weekly npm check.
PAT scopes:
GHCR_TOKENandGH_PACKAGES_TOKENshould includewrite:packagesandread:packages.GH_PAT_TOKENonly needsread:packagesfor Dependabot.
Google Calendar (OAuth2)
For Google Calendar, use OAuth2 authentication:
{
"mcpServers": {
"dav-mcp": {
"command": "npx",
"args": ["-y", "dav-mcp"],
"env": {
"AUTH_METHOD": "OAuth",
"GOOGLE_USER": "your@gmail.com",
"GOOGLE_CLIENT_ID": "your-client-id",
"GOOGLE_CLIENT_SECRET": "your-client-secret",
"GOOGLE_REFRESH_TOKEN": "your-refresh-token"
}
}
}
}
Security
- Input Validation: All inputs validated with Zod schemas before execution
- Rate Limiting: 100 requests/minute per session (HTTP mode)
- Bearer Auth: Token authentication for HTTP transport
- No Credential Storage: Pass-through only, never logged or cached
- Structured Logging: Audit trail with request IDs, no PII exposure
- CORS Protection: Whitelist origins, block cross-site attacks
Documentation
- MCP Specification - Model Context Protocol docs
- tsdav Docs - CalDAV/CardDAV library reference
- CalDAV RFC 4791 - CalDAV protocol specification
- CardDAV RFC 6352 - CardDAV protocol specification
Contributing
Pull requests are welcome! Please read CONTRIBUTING.md for guidelines.
License
MIT License - see LICENSE for details
Acknowledgments
Built with:
- tsdav - Excellent TypeScript CalDAV/CardDAV library
- tsdav-utils - Field-agnostic utility layer for RFC-compliant field updates
- MCP SDK - Model Context Protocol by Anthropic
- ical.js - RFC-compliant iCalendar parser
Questions? Issues? Create a GitHub issue
Built for AI agents managing calendars, contacts, and tasks
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。