Dashboard
The Dashboard is the app’s home page (sidebar → Dashboard). Its subtitle summarizes the moment: 2 running · 1 awaiting approval · 3 triggers watching. The New run button opens the Start a run dialog.
KPI tiles
Section titled “KPI tiles”| Tile | What it counts |
|---|---|
| Active runs | Runs that are queued, preparing or running (runs waiting for approval are counted separately). Subline: how many are queued and the global concurrency limit (2 queued · limit 3). |
| Awaiting approval | Runs stopped at a human approval gate. The tile turns amber when non-zero. |
| Success rate · 7d | Of the runs created in the last 7 days that finished as succeeded or failed, the percentage that succeeded. Cancelled runs are left out. Shows — when there are none. |
| Median lead time · 7d | Median duration of runs that succeeded in the last 7 days, from start to finish (“issue → done”). |
| Agent spend · 7d | Sum of run cost over the last 7 days, plus total tokens (input + output). Cost only includes harnesses that report it (Claude Code, Pi, the Simulator); see Harnesses overview. |
Panels
Section titled “Panels”Left column
- Needs your approval (only when something is waiting): one card per run at an approval gate. Click to open the run and approve or reject.
- In progress: active runs with their step dots, the step currently running (
Running Implement · Claude Code · sonnet) and elapsed time. All runs → opens the Runs page. - Recently finished: the last eight finished runs with status, issue, the first link found (for example
PR #57), duration and how long ago.
Right column
- Runs per day · 14d: a chart of recent run volume.
- Triggers: each trigger with its status pill (live, error, paused) and either its last error,
Checked 1m ago, orWaiting for first check. Manage → opens Triggers. - Pipelines: each pipeline with its colour, run count and the status of its latest run.
First run and warnings
Section titled “First run and warnings”With no runs yet, the Dashboard shows a welcome card with Run the demo pipeline, Connect agents (opens Harnesses) and Add a Jira / Linear trigger.
If no harness other than the Simulator is detected, a warning suggests installing one (for example npm i -g @anthropic-ai/claude-code) or adding an API key on the Harnesses page.
Live updates
Section titled “Live updates”Everything on the Dashboard updates live as runs progress, including the elapsed times. The same numbers drive the sidebar badges: Runs shows ✋ and the approval count when something is waiting, otherwise the number of active runs.