Activity & status

Bikube follows what the agent in each card is doing, so you can run many agents and only look at the ones that need you. This page explains activity and waiting, the Next waiting card shortcut, agent status lines, notifications, Claude Code's usage, and what Claude Code's hooks report by themselves.

Activity

A card's activity is what its agent says it is doing. It shows while the card is Running, as an icon before the card's title on the board and on its row in the sidebar, named in its tooltip:

A running card with no activity shows a green dot. Agents set their activity with board activity (working, waiting, question, or none to clear it); Claude Code cards set it by themselves through Bikube's hooks (see below). Activity is cleared when the card's process stops or starts again, and is never saved or synced.

Waiting

A card or shell whose activity is Waiting for you or Needs your answer is waiting. Waiting cards:

• show their bell or question mark in the sidebar tree, under their workspace;
• are counted on each workspace's badge when the sidebar is collapsed (2 waiting for you);
• come first in the grid view, in its Needs you section;
• are counted on the Dock icon on macOS, while notifications are on.

Waiting for you works like a notification: once you look at the card (you open it, or it starts waiting while you are looking at it), it counts as seen and clears. Needs your answer stays until the agent hears back from you.

Next waiting card

⌘J / Ctrl+Shift+J (default shortcut) opens the card that has been waiting longest, in whatever workspace, switching to that workspace if needed. Press it again for the next one: if the card you are looking at is waiting, it goes to the one after it in the queue. When nothing waits, nothing happens. It is also in the command palette as Next waiting card.

With a global shortcut set, bringing Bikube forward from any app can open the next waiting card too, so you can jump between agents without hunting for them.

Agent status

A card's agent status is a short line its agent writes about its progress with board status, such as Tests green, PR ready for review. It shows under the card's title while the card runs, up to two lines (the full text in its tooltip), above the last output line. An empty text clears it. When the card stops, its last status stays on the card, dimmed.

Claude Code cards keep one by themselves: Bikube's hooks ask Claude to write a one-line status at the end of every turn. Turn this off with Claude Code status in Settings, Agents (on by default); it applies to cards started from then on.

Agent status is separate from the card's status (Running, Stopped, SSH failed, Couldn't start), which Bikube works out itself; see Cards.

Notifications

With Notifications on in Settings, Agents (on by default), Bikube notifies you when a card's activity becomes Waiting for you or Needs your answer, unless you are looking at that card. The notification has the card's title and, for example, Needs your answer · api; clicking it opens the card (or brings its card window forward). Notification sound plays a short tone with each one (on by default).

Notifications also tell you when a command that ran 10 seconds or more ends in a terminal you aren't looking at (see Split panes & terminal features), and when a trigger fails.

Usage

A local Claude Code card shows how much of its conversation's context window is used, as a bar and a percentage under its status lines (42% context). Hover it for the details: the context used of the window, then the input, output, cache write and cache read tokens, and the model. No cost is estimated.

Bikube's hooks report the usage with board usage from Claude Code's transcript when a session starts and after every turn. The bar stays after the card stops, and goes when the conversation is cleared with /clear or the card is deleted. Only local Claude Code cards have it: not other agents, and not SSH cards.

What Claude Code's hooks do

Every Claude Code card starts Claude with Bikube's own settings file (--settings "$BIKUBE_CLAUDE_SETTINGS"), which adds hooks and nothing else. Each hook runs a quiet board command for the card and never fails or holds Claude up:

The instructions Claude receives at the start of a session tell it that it runs inside a Bikube card and to run "$BIKUBE_BIN/board" status . '<one line>' at the end of every turn, silently, and that the card may have a task description to read with board description .. In local cards they also tell Claude how to read the board and add cards when you ask (board list --json, board create), and how to port work to another workspace (board workspaces, then board create --workspace … --description-file -, with --worktree when you ask for one). You can see Bikube's hooks on the Agent setup page, with the Bikube source.

Other agents don't get hooks. They can still drive their card with the board command, if you or their instructions ask them to.