TickTick MCP
A remote MCP server that enables Claude to create and manage TickTick to-dos using the TickTick Open API. It supports nine tools for projects, tasks, and sections, and works across all Claude platforms (web, mobile, desktop, Cowork).
README
TickTick MCP — remote connector for Claude
A single-user remote MCP server that lets Claude create and manage TickTick to-dos. Because it's a remote connector (Claude reaches it over HTTPS from Anthropic's cloud), one deployment works on all platforms: claude.ai web, the mobile apps, Claude Desktop, and Cowork.
It wraps the TickTick Open API and exposes nine tools:
| Tool | What it does |
|---|---|
list_projects |
List your projects/lists (call first to get a projectId) |
list_sections |
List a kanban project's sections/columns (to get a columnId) |
list_tasks |
List open tasks in a project (with each task's section) |
get_task |
Full detail of one task |
create_task |
Create a to-do (Inbox, a project, or a kanban section) |
update_task |
Edit fields on a task (incl. moving it between sections) |
complete_task |
Mark a task done |
delete_task |
Permanently delete a task |
create_project |
Create a new list/project |
Dates are handled for Africa/Harare (UTC+2) — pass YYYY-MM-DD (all-day) or YYYY-MM-DDTHH:MM (timed) and they land on the day you mean.
Setup (about 20 minutes, one time)
1. Register a TickTick app
Go to https://developer.ticktick.com, create an app, and set its redirect URI to:
http://localhost:8080/callback
Note the Client ID and Client Secret.
2. Get an access token
npm install
npm run auth
Paste your Client ID/Secret when prompted, authorize in the browser, and the script prints your TICKTICK_ACCESS_TOKEN. TickTick issues a long-lived token (~180 days) and no refresh token, so re-run npm run auth and update the env var when it expires.
3. Generate a URL secret
node -e "console.log(require('crypto').randomBytes(24).toString('hex'))"
This becomes the secret path segment in your connector URL. Treat it like a password.
4. Test locally first (optional but recommended)
You don't need the Vercel CLI to test. Create .env.local (copy .env.example) with all four vars, then:
npm start
This serves the exact production handler on http://localhost:8080. Connect Claude Code:
claude mcp add --transport http ticktick http://localhost:8080/api/<MCP_SHARED_SECRET>/mcp
Then ask Claude Code to list your TickTick projects to confirm it works end to end.
5. Deploy to Vercel
Create a new Vercel project from this folder, then add these environment variables (Project → Settings → Environment Variables):
TICKTICK_CLIENT_IDTICKTICK_CLIENT_SECRETTICKTICK_ACCESS_TOKENMCP_SHARED_SECRET(the secret from step 3)
npm run deploy
Your endpoint is:
https://<your-app>.vercel.app/api/<MCP_SHARED_SECRET>/mcp
6. Add it to Claude
Settings → Connectors → Add custom connector → paste the full URL above → Add. Leave the OAuth fields blank (this server is authless; the secret in the URL is the guard). Once added on one platform it's available on all of them.
How it stays secure for a single user
There's no OAuth handshake between Claude and this server. Instead the secret lives in the URL path, and the server returns 404 to any request whose path secret doesn't match MCP_SHARED_SECRET. Your TickTick tokens never leave the server — Claude only ever calls the eight tools.
If you later want to share this with teammates (each hitting their own TickTick account), that's the point where you'd swap the path-secret model for a real OAuth flow. The tool logic in lib/ticktick.js wouldn't need to change.
Notes
- The server is stateless per request, which suits serverless — no session state to lose between cold starts.
- The TickTick access token is long-lived (~180 days) and used directly — TickTick's Open API doesn't issue refresh tokens, so there's no auto-refresh. Re-run
npm run authand update the env var when it expires. delete_taskis irreversible; the tool description nudges Claude towardcomplete_taskunless you clearly want deletion.
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。