Your AI can write code, summarize meetings, and research anything. But can it manage your to-do list? With TickTick MCP, it can — your AI assistant becomes a task manager you talk to instead of a UI you click.
The problem
Every task manager has a UI, and every AI has a chat box. The friction is the gap between them: you think of something, you have to switch to the app, click through to create a task, set the date, pick the list, tag it. That’s exactly the kind of repetitive friction that AI should be eating.
TickTick MCP closes that gap. It’s a Model Context Protocol server that lets MCP-compatible AI clients talk to your TickTick tasks directly — read them, create them, update them, complete them — all from the chat interface.
What it actually does
"What tasks do I have today?"
"Create a task: Prepare meeting materials tomorrow at 3:00 PM, high priority."
"Mark 'Buy milk' as completed."
Your AI parses that natural language, calls the right TickTick MCP tool, and the task is created — with the date, time, priority, and list extracted automatically. No app switching, no manual entry.
The server URL is https://mcp.ticktick.com, using the Streamable HTTP transport with OAuth (or a Bearer Token).
The tool catalog (what your AI can do)
TickTick MCP exposes a full set of tools, grouped by capability:
Task queries — search tasks, get by ID, list undone/completed by date range, filter by list/priority/tag/status.
List management — list/create/update projects (lists), sections, and folders.
Task management — create, batch-add, complete, update, move, and delete tasks; comments; assignment; tags.
Habits — list, create, update habits and their check-ins.
Focus records — query, create, and delete focus sessions.
Countdowns — list them.
The nice part: you don’t need to memorize tool names. Describe what you want in natural language and the AI picks the tool.
Setup by client
Claude Code
claude mcp add --transport http ticktick https://mcp.ticktick.com/
Then run /mcp in the session and complete OAuth. For a Bearer Token:
claude mcp add --transport http ticktick https://mcp.ticktick.com/ --header "Authorization: Bearer YOUR_TOKEN_HERE"
Cursor
Edit .cursor/mcp.json:
{
"mcpServers": {
"ticktick": { "url": "https://mcp.ticktick.com" }
}
}
Save, reopen Tools & MCP, click Connect, complete OAuth.
VS Code
Edit .vscode/mcp.json:
{
"servers": {
"ticktick": { "type": "http", "url": "https://mcp.ticktick.com" }
}
}
Or Command Palette → Add Server → HTTP → enter the URL.
Codex
codex mcp add ticktick --url https://mcp.ticktick.com
OAuth completes in your browser automatically.
Claude, ChatGPT, Gemini
Each has a built-in connector flow — Claude (Settings → Connectors), ChatGPT (Plugins), Gemini (Connected Apps) — search “TickTick” and follow the OAuth prompts.
The CLI, for terminal-first people
If you’d rather skip the MCP layer and just script it, TickTick also ships a CLI:
npm install -g @ticktick/ticktick-cli
ticktick auth login # browser OAuth
ticktick project list # view lists
ticktick task create --title "Task name" --project <list ID>
ticktick task complete <list ID> <task ID>
Same task management, from the terminal — which makes it trivial to drop into cron jobs or agent scripts.
Why it matters
The whole point of an AI agent is removing friction between intent and action. TickTick MCP is a clean example: the intent (“remind me to reply to that email”) becomes an action (a dated, prioritized task in the right list) without you touching the app. For anyone living in an AI chat interface all day, that’s a meaningful quality-of-life upgrade.
And it’s a good reminder of what MCP is really for — not just letting AI read your data, but letting it act on it. The task manager becomes something you talk to, not something you click.
Next step
Connect it to your favorite client, then try the natural-language flow: “What tasks do I have today?” → “Create a task: …” → “Complete ‘Buy milk’.” Once you feel the friction disappear, you won’t go back to clicking.