Executions
The Executions page is the run history: every run of every automation, newest first, with each run's steps, logs, and output one click away. It is a record, not a control panel—you read runs here and open them from here.
The same page also works narrowed to a single automation. Clicking an automation's Runs count on the Automations list, or Runs from inside its editor, opens this page scoped to that automation: the breadcrumb reads Executions › automation name, an Edit Live link appears for jumping back to the editor, and the Automation column drops out because every row shares it.
What the table lists
| Column | What it shows |
|---|---|
| Automation | Which automation the run belongs to, linking to that automation's own run history. Shown only when the list covers every automation. A run outlives the automation that produced it, so a run whose automation has been deleted shows the last name it had, unlinked and muted. |
| Run | The run number, as #42, with a Dry Run badge beside it when the run was a dry run. A live run gets no badge. |
| Status | Running, Complete, Errored, or Cancelled. Everything a run passes through on its way to finishing reads as Running. |
| View Execution | A pair of buttons, List and Workflow, opening the run's two views—its steps as a list, or the flow on a read-only canvas. |
| Started At | When the run began, as a relative label such as "15 hours ago" unless full timestamps are on. |
| Duration | How long the run took. With full timestamps on, the cell also carries when the run ended; a run still going shows only its duration so far. |
| Cost | The approximate cost of this run's AI work, calculated from the tokens it used, to four decimal places, or - when there was none. |
| Trigger | What started the run, named as the canvas names it, followed by what that trigger has to say about this particular run—for example "On Demand - Nightly kickoff · Triggered by Ada Lovelace". Both halves are read from the automation as it was when the run happened, so the row still reads correctly after the automation is edited. |
| Logs | A View Logs link, opening the run's log viewer in a popup over the page. |
A run that is still going pulses in the list, and runs that start while you have the page open arrive at the top on their own—no reload needed.
Controls
- Status—filter to All Status, Running, Complete, Errored, or Cancelled.
- Full timestamps—a switch that expands Started At and Duration from relative labels to exact dates and times. It is remembered for you and survives paging.
- Paging—first and previous links with a page-size selector. Changing a filter keeps your place in the list rather than throwing you back to page one; if the narrowed filter has fewer pages than your current position, you land on page one instead of an empty page.
- View Older Executions—appears at the foot of the last page when there is archived history beyond the retention limit. Following it continues the same list with archived runs included.
When nothing has ever run, the page shows No Executions Yet in place of the table.
The log viewer
View Logs opens one run's log: its steps' status transitions merged with the messages the steps produced, in time order. Opening it from a single step instead narrows it to that step, and the heading names which one.
| Column | What it shows |
|---|---|
| Timestamp | To the millisecond, in your profile's time zone. |
| Type and Node | Which step produced the entry, by node type and by its label, with the node's category icon beside them. |
| Iteration | Which pass of a loop the entry came from, named after the loop—for example "slides:0 / tags:1". Empty for a step outside any loop. |
| Level | System, Debug, Info, Warn, or Error. |
| Message | The entry itself. A long or multi-line message is truncated to its first line; click it to expand and click again to collapse. |
The type, node, and iteration columns are dropped when the log is already narrowed to one step.
Filter the level with the buttons above the table: All, System, Debug, Info, Warn, Error. The filter is applied to the whole run rather than to the page on screen, so filtering to Error finds every error in the run, not just the ones that happened to be on the page you were reading. Status transitions are System entries, but a failed step's error message is an Error one, as is the closing line of a run that failed—so Error finds failures without hiding them. Filtering to a level that matched nothing says so, names the level, states how many entries the run did produce, and offers Show all rows to undo it.
Debug entries are recorded only when Capture Debug Logs is turned on from the Settings page. It is off by default; turn it on while working out why an automation is misbehaving, and off again afterwards.
How the log pages depends on whether the run is still going.
- A finished run gets numbered pages with a page-size selector.
- A live run gets Load earlier 50 (or whatever the page size is) and Jump to newest, and opens on the newest entries—entries arriving while you read never renumber the page under you. A run that finishes while you are reading it switches to numbered pages.
A run long enough to exceed what one view can hold shows a banner above the table saying how many entries are on screen out of how many there are, so a trimmed log is never mistaken for a complete one. When no level filter is on, the banner also suggests filtering by level to narrow the run instead.
See also
- Monitoring and history—how to watch a run unfold, read its steps and output, and work out why one failed.
- Dashboard—the summary view of the same activity.
- Human in the loop—why a run may be sitting in Waiting rather than progressing.