One board for all your coding agents

The manual way

# one list per tool, per account, per computer
claude --resume
CLAUDE_CONFIG_DIR=~/.claude-work claude --resume
codex resume
ssh studio -t 'codex resume'

Each of those answers for one harness on one account on one computer. Who is waiting on you across all of them is something you work out by going through the windows.

Add the Agent board

1. Start wisp.
   npxnpmcurlbrewDownload

   $npx wisp-agentsCopy

   macOS · Linux and Windows in beta · Node 20+ · MIT

2. Open Make a pane. ⌘K → Make a pane…, or Make a pane at the top of My Stuff. Press Add on Agent board.
3. Say what it may use. Wisp’s card lists it. Allow and pin puts the board on home, under Needs you.

No agent runs and no tokens go: Agent board ships with wisp, and it reads the same task list as the sidebar.

What’s on the board

• Working: the agent is running. Needs you: it asked you something or is waiting. Review: it finished and you haven’t marked it done. Done: you have.
• Each card shows the title, the harness, the account when a harness has several, the machine when it isn’t this computer, the PR and how long ago it moved.
• Your crew sits above the columns, with who’s working now. Their jobs are tasks, so they are cards like any other.
• Done: drag a card to the column or press its button, by mouse or finger. It works on tasks on other machines too.
• Follow up opens wisp’s own card with the task, harness, mode and your exact words. Nothing goes until you press Send.

The board updates as tasks change, and a card you’re typing in stays open while it does.

What other tools’ panes and boards do

Pinned panes are no longer rare. The difference is what a pane can see.

The first four stay inside their own harness. Vibe Kanban and Cline Kanban cross harnesses for work you start from their board. wisp’s board reads wisp’s task list, so whatever wisp runs is on it: a crew member’s check-in, a task on your GPU box, a run on your second account.

Allowance board: every account’s limits, and what each period cost

Add it the same way. It keeps a card per account open on home: how much of each 5-hour and weekly window is gone, when it resets, and how fast it’s going once that can be told.

Under each window sit its recent periods, newest first: how much went outside wisp, how much was used, and what 1% of the window cost in wisp, in dollars on Claude Code and in millions of tokens on Codex. Outside wisp can read low, never high. A cost worked out from running totals is marked ≈.

Claude Code and Codex report their windows; the other harnesses say so on their card. It’s read-only: the board can’t start or change anything. For which tasks used a window, see where your usage went.

wisp grows with you

Both boards are ordinary apps in My Stuff. Want a column for PRs with failing checks, or one board per project? Type make me a pane that… in New task and an agent builds it in an empty folder, pins it, and the pane reloads in place as it changes. Claude Code mods can be made from a prompt too; what a wisp pane adds is that it can see all of your work.

A pane gets only what you allow, when you press Allow… on its bar:

• It reads each task’s title, state, project, branch, PR and the last answer’s one-line summary. Never the conversation, and secrets are redacted first.
• It can’t send a prompt. Every task it starts and every follow-up waits on wisp’s card for your click, at most 6 writes a minute, in your default harness and mode.
• If its code changes, its write access stops until you allow it again.