Dashboard

The Dashboard is where you land after signing in. It answers one question at a glance: is everything running the way it should be?

Dashboard

The stat row

Four cards summarise the last 30 days. Each one is also a control — click a card and the report behind it opens in the panel below, so the number and the detail that explains it are never more than one click apart.

Card What it counts Opens
Steps w/ Errors Failed step executions in the last 30 days Runs, filtered to failures
Executed Deployments Pipeline runs in the last 30 days Pipeline Executions
Total Tokens Consumed AI tokens used across all agent calls Audit Trail
Transfer Usage Pipeline data stored plus files ingested Transfer Usage report

All four use the same 30-day window so the figures can be read against each other.

The panel

Below the cards, the selected report renders in place. It opens on Steps w/ Errors — failures are the thing worth seeing first — and each report also has a full-page version, linked from the panel header.

That default report is Runs, pre-filtered to failures over the same 30 days, so you land on what broke rather than on a list of everything.

To manage deployments themselves — pausing, rescheduling, changing a pinned version — go to Deployments in the sidebar. The Dashboard panels are read-only views.

Getting started callout

Until your organization has a connection and a data step, a short callout appears above the stat row pointing at the next thing to do. It disappears on its own once you're set up.

A new account looks empty, and should

Every card counts the last 30 days. On a fresh organization they read zero, and that is the correct answer rather than a loading failure — there is nothing to count yet. They start moving once a pipeline has run.

If a figure looks wrong on an account that has been running, the usual causes are mundane:

  • Steps w/ Errors is zero but you know something failed. The window is 30 days and the count is of step executions. A pipeline that never started — a paused deployment, a schedule that did not fire — has no failed step to count. Check Pipeline Executions instead.
  • Total Tokens looks high for the chat you remember. It counts every agent call, including the schema enrichment that runs when you index a source. Only some of that is billable; the usage report splits the two.
  • Transfer Usage is informational. Going over your allowance is not blocked and nothing stops running.