Kanban board
The Board is a kanban view of the project's work items. Its special trick: cards can be assigned to AI agents, and dragging an agent's card to To Do kicks off an agent run.
Every card is also a task of the project's plan: the Gantt page draws the cards on a timeline, with start dates, sub-tasks and dependencies. See The plan (Gantt).
A card can carry To-dos: open the card, type one under To-dos and press Add, and tick it off when it is done. They are what fills the task's bar on the Gantt, and the To-dos page lists them with the project's others.
Columns
Backlog · To Do · In Progress · Review · Done. Each column shows its card count and has a + Add card composer at the bottom.
Creating a card
The composer takes:
- Title (required) and an optional description.
- Assignee — Unassigned, Me, or any agent.
- Artifact IDs — optional comma-separated artifact IDs to attach the relevant requirements to the card.
Driving agents from the board
Tip shown on the board itself: assign a card to an agent and drag it to To Do to launch the agent.
- When you move a card assigned to an agent into To Do, an agent run is enqueued for that card (skipped if a live run is already attached). Only human moves trigger this — the automatic card moves that track run progress don't.
- A pulsing blue dot on a card means a run is live (queued, claimed, or running). While a run is live the card is locked — it can't be dragged until the run finishes.
- The agent's progress and comments land back on the card's activity feed.
- A chat is not a task. Talking to the assistant, or typing a prompt to an agent with ▶ Launch run, puts no card on the board: the run is on the Runs page. A card is something you assigned; an automation or a schedule that starts an agent on a project still gets one, so its work is tracked.
The card drawer
On a computer, click a card to open it beside the board: the board gives it room and stays in view, so you can go from card to card. The open mark at a card's corner, or in the open card's header, opens the card in full, wide in the middle of the window. On a phone a card opens over the board. Either way the card lets you:
- Edit the title and description.
- See linked artifacts (from the artifact IDs) with links into the Requirements view.
- Follow its History — comments, moves, status changes, and agent run events — and add your own comments. It is the same History an artifact has: Comments, Changes and All choose what it lists, and with the card opened in full it is a card of its own.
- Delete the card.
Opened in full on a computer, the card also shows everything the plan holds on it, on the right: its schedule (start, due date, duration, effort and deadline), the task it is a sub-task of, who is assigned and what has been done, its milestone, its relationships with other tasks, its to-dos and its sub-tasks. It is the same panel the Gantt page shows for the task, and a change made in either place shows in both: see The plan (Gantt). Selecting a sub-task or a related task there opens that card.
Moving cards
Drag any unlocked card between columns or within a column. Human-assigned and unassigned cards behave like a normal kanban; only agent cards with live runs are locked.
A card that was assigned to a crew, before crews were removed, is now assigned to the crew's first (entry) agent, or is unassigned if the crew had none.
To-dos: small jobs, kept off the board
A to-do is one paragraph of text: small enough that you just do it and tick it off, like check the wording of REQ-12 or ask Sam about the seal supplier. It has no column, no discussion thread and no agent runs — those are what a board card is for. The board shows only cards, so a list of small asks no longer crowds out the real pieces of work.
A to-do has its text, a tick box (and when it was ticked), who has it, an optional due date, and, when it was raised from a note, a link back to the note's artifact.
The project's to-do list
Plan → To-dos is the project's list:
- Quick add at the top: type the to-do and press Enter. Choose who has it and a due date beside the field if you want to; both are optional.
- Tick the box to complete a to-do. Completed to-dos collapse into a Completed line under the list; open it to see them, or untick one to open it again.
- Edit changes the text, who has it and the due date. Anything overdue is called out.
- Just mine narrows the list to the to-dos assigned to you.
- A to-do is one paragraph, at most 1000 characters. Anything longer is a piece of work: put it on the board.
My to-dos
My to-dos, the To-dos tab under Plan in the workspace's rail (and linked from every project's To-dos page), gathers your to-dos across the workspace: every to-do assigned to you in a project you can open, each labelled with its project, and your personal to-dos. A personal to-do belongs to no project and nobody else sees it; the quick add on My to-dos makes one.
Promote to board task
When a to-do turns out to be bigger than it looked, Promote to board task on its row makes a card of it: the card lands in To Do carrying the to-do's text, who has it, its due date and the artifact it came from. The to-do is ticked off with the note promoted to a board task, which links to the card.
Raising a to-do from a note
Mentioning someone with @name in a note on an artifact tells them about it — it does not create work for them. To turn the ask into something tracked, use + Add to-do on the note:
- Whoever the note names is filled in as the person who has it, and the note's first paragraph as the to-do. Both can be changed, and you can set a due date.
- The to-do goes on the project's to-do list, and on the My to-dos of the person who has it, straight away. It does not go on the board.
Afterwards the note carries a small link to the to-do showing whether it is open or done, so a thread from three weeks ago tells you whether the thing being discussed ever got done, without going to look. The status is read from the to-do each time, not copied.
A note raised before to-dos had a list of their own keeps its link to the board card it raised, and the card stays on the board.
A note can raise one to-do. System and agent entries in the feed do not offer the control: those are a record of what happened, not somebody asking for something.
For agents
Agents reach to-dos through the openv tools list_todos, create_todo
and tick_todo, so a run can leave a small follow-up for a person, or
tick one off once it is done.