Every cnvs.app board can be a kanban to-do board instead of an infinite canvas. Switch the mode while the board is empty and you get columns, optional lanes, and cards carrying priority, due date, assignee and a done state — all syncing live to everyone with the link. No account, no install, and a public MCP endpoint so AI agents can manage tasks alongside you.
A board is either draw (the infinite canvas) or todo (a kanban task board). You pick the mode with a single button while the board is still empty — once a board has content the mode is fixed, so a drawing never silently turns into a task list. In to-do mode the board becomes:
| Field | What it's for |
|---|---|
| Name | The task title — the one required field. |
| Description | Optional longer detail / notes on the card. |
| Priority | Flag how urgent a task is, set inline while adding. |
| Due date | When it's due — set inline while adding. |
| Assignee | Who owns the task — set inline while adding. |
| Done | Mark a card complete without deleting it. |
You can paste a plain list of lines and turn each line into its own task, and delete a hovered card straight from the keyboard (Del / Backspace).
Cards, columns and lanes sync over the same real-time WebSocket layer as the drawing canvas. Open the board on two devices, add a card on one, and it appears on the other within a moment — no refresh. There are no accounts and no tracking: the board ID in the URL is the only key. Share the link and everyone can edit. More on real-time collaboration →
cnvs.app is an installable PWA. Every board you've opened is cached for offline read-only access, so you can review your task board on a flaky connection — edits sync back when you're online again. More on offline / PWA →
The same public Model Context Protocol endpoint at https://cnvs.app/mcp that powers AI on the drawing canvas also manages tasks. An MCP-aware AI like Claude can create, move, complete and list cards on a shared to-do board:
set_board_mode — switch an empty board to todo mode
create_task — add a card (name, description, priority, due, assignee)
move_task — reorder / move a card between columns and lanes
list_tasks — read the current board as structured JSON
Every tool has a 1:1 REST mirror — POST /api/boards/<id>/mode, POST/DELETE /api/boards/<id>/columns, POST/DELETE /api/boards/<id>/tasks (+ /tasks/<id>/move) — and the subscribable resource cnvs://board/<id>/tasks.json so an agent gets a push when the board changes. See the developer docs →
A to-do board exports to PDF straight from the browser — handy for a printed standup snapshot or an attachment — and the structured board state is available as JSON over the API for any downstream tooling.
Yes — every board is free with no signup, including unlimited columns and cards with real-time multi-user sync.
Only while the board is empty. The mode (draw vs todo) is fixed once a board has content, so a drawing never turns into a task list by accident. Open a fresh board to start a to-do board.
Yes. Dragging cards has full touch support, so reordering works on phones and tablets, and the board is installable as a PWA.
Yes — via the public MCP endpoint and its REST mirror, an agent can create, move, complete and list cards alongside human collaborators. More on AI via MCP →