A year ago, “connect my to-do list to my AI” meant gluing together a community script and hoping it didn’t break. In 2026 it has a name — MCP, the Model Context Protocol — and a growing list of task managers that support it. MCP is an open standard that lets AI clients like Claude, Cursor, and Codex talk to your tools, so your assistant can read and write your actual tasks instead of you describing them by hand.
Here’s an honest rundown of where MCP support actually stands across the task apps people ask about.
TickTick — first big incumbent in
TickTick added MCP support in a 2026 update, which made it the first mainstream to-do app to ship it. If you already live in TickTick and just want Claude to add the occasional task, it’s a reasonable place to be. The caveats: it’s a feature on top of a sprawling app (habits, Pomodoro, calendar, Eisenhower matrix), and the design isn’t for everyone.
Todoist — community servers, not first-party
Todoist itself doesn’t ship an MCP server, but its mature REST API means the community has built several (search GitHub for “todoist mcp server”). They work, and Todoist’s API is solid. The trade-off is that you’re running and trusting a third-party bridge, keeping it updated, and managing your own tokens. Fine for tinkerers; not turnkey.
Linear — great if your “tasks” are issues
Linear has strong MCP support, but it’s an engineering issue tracker, not a personal to-do app. If the things you want your AI to manage are tickets and projects, it’s excellent. For “buy milk, call the dentist, draft the proposal,” it’s the wrong shape.
Lume — first-party MCP plus agent hand-off
Lume ships a first-party MCP server built for personal tasks, in a Things-style design. The difference isn’t just that the MCP is native and per-user — it’s what sits next to it: you can also assign tasks to Claude Code or Codex and have them do the work, plus a REST API for scripting. It’s the only one on this list where the AI can both manage the list and clear items off it. (Disclosure: I build it.)
How to choose
- Already in TickTick and happy? Turn on its MCP and move on.
- Live in Todoist and comfortable running a script? A community MCP server will do the job.
- Your tasks are engineering issues? Linear.
- Want a clean personal list your AI can manage — and that can hand work to agents? That’s the gap Lume is built for.
The bar is rising fast
Two years ago none of these existed. Today MCP is going from novelty to expectation — the same way calendar sync and natural-language dates did before it. If you’re picking a task manager now, “can my AI reach it?” is a fair question to ask, and increasingly the answer is yes. The more interesting question is the next one: once your AI can see your list, do you also want it to do the list?
If that second part appeals, start with the MCP task manager overview, or read how connecting your list to Claude actually works.