dashboard: global agent kanban view — Working / Waiting / Completed / Error / Abandoned #977

New issue

Closed

opened 2026-05-08 12:17:44 +00:00 by claude-desktop · 1 comment

claude-desktop commented

2026-05-08 12:17:44 +00:00

Collaborator

User story

As an operator overseeing the full fleet, I want a kanban view across all active and recent agent runs grouped by status (Working / Waiting / Completed / Error / Abandoned), so the at-a-glance "what's the state of my fleet" question takes one screen instead of two pages of clicking.

Context

The Claude Code Agent Monitor uses exactly this columnar layout (repo) and it's the highest-density way to take in a fleet. Different from our existing planner/board (which is per-issue) — this is per-task-run.

Acceptance criteria

New route /agents/fleet (or pull into existing /monitor).
Five columns: Working, Waiting, Completed (last 24h), Error (last 24h), Abandoned (last 24h). Counts in the column header.
Each card: agent type, model, task title, repo/issue link, status pill (with reason on hover), duration, token meter, cost.
Click card opens the existing run drawer.
SSE live updates re-shuffle cards into the right column.
Filter chips: agent type, repo, provider.

Out of scope

Drag-to-reroute between columns (not a meaningful operation here).
Per-card cancel/pause (those live on the drawer; the kanban is a navigator).

Dependencies

Depends on status-pill-vocabulary ticket (column derivation).

References

Claude Code Agent Monitor: https://github.com/hoangsonww/Claude-Code-Agent-Monitor

## User story As an operator overseeing the full fleet, I want a kanban view across all active and recent agent runs grouped by status (Working / Waiting / Completed / Error / Abandoned), so the at-a-glance "what's the state of my fleet" question takes one screen instead of two pages of clicking. ## Context The Claude Code Agent Monitor uses exactly this columnar layout ([repo](https://github.com/hoangsonww/Claude-Code-Agent-Monitor)) and it's the highest-density way to take in a fleet. Different from our existing planner/board (which is per-issue) — this is per-task-run. ## Acceptance criteria - [ ] New route `/agents/fleet` (or pull into existing `/monitor`). - [ ] Five columns: Working, Waiting, Completed (last 24h), Error (last 24h), Abandoned (last 24h). Counts in the column header. - [ ] Each card: agent type, model, task title, repo/issue link, status pill (with reason on hover), duration, token meter, cost. - [ ] Click card opens the existing run drawer. - [ ] SSE live updates re-shuffle cards into the right column. - [ ] Filter chips: agent type, repo, provider. ## Out of scope - Drag-to-reroute between columns (not a meaningful operation here). - Per-card cancel/pause (those live on the drawer; the kanban is a navigator). ## Dependencies - Depends on status-pill-vocabulary ticket (column derivation). ## References - Claude Code Agent Monitor: https://github.com/hoangsonww/Claude-Code-Agent-Monitor

claude-desktop added the

area:dashboard

type:user-story

labels

2026-05-08 12:18:19 +00:00

claude-desktop added a new dependency

2026-05-08 12:19:19 +00:00

#963 dashboard: status pill vocabulary — Planning · Executing · Waiting · Reviewing · Done · Error · Abandoned (with reason hover)