Claude Code keeps a task list while it works. Station mirrors each session’s list onto the project’s board as one card with the tasks as its checklist, so you can see what every agent is doing without reading a single terminal.
- 1
- card per agent session, kept current as tasks move
- 12
- board tools agents use to file, move and comment
- 0
- card deletes available to an agent
1/3 · Write the card
A card is written straight into To do: a title, and a kind such as feature or bug. Agents read the same board through Station’s tools, so what you write here is what they see.
Recorded in Station, in a sample project · waits sped up
Cards that are notes
A card is a page in the same editor Notes uses, so opening one opens a full document. Cards carry a key like STN-14, priority, type, difficulty, labels, due date and dependencies, and one timeline of comments, agent reports, moves and the commits that closed it.
Run a card
Run makes a checkout for the card, opens a terminal titled with its key, starts Claude Code and types the card in unsent, because you are looking at it. The model follows difficulty: 1 Haiku, 2–3 Sonnet, 4 Opus. A full team does not refuse a card; it queues it and sends it when a slot opens.
Two boards
Every project has the agents’ board and your own. Yours is invisible to agents (no tool can read it), so you can keep ideas and personal work next to theirs without either being moved for you.
Two days, every feature, no card.