A todo list with Claude built in: why I keep coming back to this layout

every todo app added "ai" in 2024. most of them did it the wrong way: a sidebar with a generic chatbox that can't see your todos, plus a few canned actions like "summarize my list" that nobody uses twice.

taskelo is a take on a different layout. one page, three panels: todos on the left, a notes canvas in the middle, claude on the right. the chat panel already knows what's on your list — every message you send includes your full todos and notes as context.

what changes when the model sees your list

three things, none of them dramatic on their own:

"plan today" stops being a useless question. the model can read what's open, what's stale, what's recurring, and give you a real ordering. if you've got a 9am call and you wrote "draft proposal" three times this week, it will tell you to do the proposal first.

you can ask half-formed questions. "i'm stuck on the third one" is a complete sentence to taskelo's chat panel — it can see what the third todo is. you don't have to re-state the context every time, and you don't have to copy-paste a list into the prompt.

you stop opening claude.ai. the friction is gone. if claude is already in the page with your work loaded, the friction of opening a separate tab feels disproportionate.

the layout

none of this syncs to a taskelo server. it lives in your browser. you can export everything as json (one button), or wipe it (one button).

what the chat actually does

a few examples of how the layout pays off in practice.

"break the third one into subtasks"

the model reads your list, picks the third item, and offers 4-6 subtasks. you can ask it to push them into the todo list directly via a sample-prompt button, or you can copy them in by hand.

"i wrote some notes about the database migration in the canvas, what am i missing"

the canvas content is in the prompt. claude can spot gaps without you having to paste anything.

"what did i finish yesterday"

the list includes timestamps. the model can group completed items by date and give you a summary. useful for standups and weeklies.

"draft a friday email summarizing the week"

both completed and open todos plus the canvas notes — that's enough material for a coherent weekly recap. the draft lands in the chat, copy it out, send it.

why this is not a deep ai integration

taskelo doesn't:

• generate todos for you automatically.
• background-classify your tasks into projects, priorities, or tags.
• learn over time. the model state is whatever's in the current conversation; there is no embeddings db, no fine-tune, no rag.

(the one proactive exception: a daily brief, on the hosted plan, that emails or pushes you a short summary at a time you pick — see the daily brief post for how that works and why it's kept separate from the chat context.)

this is on purpose. those features sound great in demos and are usually wrong in practice — they create cleanup work, they're hard to undo, and they make a simple tool feel like an enterprise product. taskelo's bet is that "model can read your list" is the only ai feature you actually need.

getting started

taskelo.com. type into the todo list. write in the canvas. open the chat panel on the right.

for the chat to work, you either:

• paste an anthropic api key (free, billed directly by anthropic)
• sign in with a magic link and subscribe for $8/month hosted

either way, the same chat ui, the same todos-and-notes-as-context behavior. no signup is required for the todos and notes alone — they work without any of the chat setup.

if you've been bouncing between a todo app and claude.ai for a year, this layout is worth fifteen minutes.