Understanding your runs
Your workspace room thread is where every run happens. When you open a team in your workspace, you'll see a chat-style thread — this is where runs appear, gates surface for review, and you can ask questions about results.
What is a run?
Each time your team executes, a run card appears inline in the room thread. It shows the overall status of the run, a token count, and — once complete — a per-agent breakdown and an output preview.
Runs work the same whether you trigger them manually, on a schedule, or via a test. A test run uses mock data so you can see how the team behaves without consuming real credentials or sending anything live.
Triggering a run
Manual run — Press Shift+R in the room thread input, or type something like "run it" and the team will pick it up.
Test run — Click the flask icon in the input bar to switch to test mode, then run as normal. The run card will show a TEST badge and the output will note that mock data was used. Useful for verifying your team is working before pointing it at live data.
Scheduled run — If a schedule is configured, runs fire automatically at each interval. You don't need to do anything — the run card will appear in the thread when it fires.
Reading a run card
While a run is in progress the card shows Running… with an elapsed timer and live token count. Once it finishes, the status updates to one of:
Run completed — Everything ran and finished successfully. An output preview appears beneath the card. Click the expand icon to read the full output.
First run completed — Same as above, shown on your team's first successful run.
Partially completed — One or more agents finished but at least one failed. The per-agent breakdown beneath the card shows which agents completed (green dot) and which failed (red dot) along with any error message.
Run failed — The run could not complete. The breakdown shows where it failed. If credentials are the cause, an Update credentials button appears. Otherwise a Fix this → link is shown.
Connection lost — The live connection to the run was interrupted. Use the Retry button to re-trigger.
After any completed run, a Run again button and an Ask about results button appear at the bottom of the card.
Human gate queues
When a run reaches a human gate, it pauses and an approval card appears inline in the thread. The card is highlighted in amber and shows the gate label, the content waiting for review, and an Approve & continue button.
Click Approve & continue to release the gate and let the run proceed. If you need to read a long output before deciding, click the maximize icon in the gate card header to open it in a fullscreen reader.
Once approved, the run resumes automatically and the result is preserved in the thread below the gate card.
Cancelling a run
While a run or a response is streaming, a Cancel button appears next to the activity indicator in the input bar. Click it to stop the current stream. Cancellation is best-effort — if the active step has already completed by the time you cancel, the run finishes normally.
Run history
Open the Observability tab in your workspace to see a full log of every run for each of your teams — completion status, duration, token counts, trigger type, and any error detail. Click any row to expand the full per-agent breakdown and output for that run.
Workspace documents
For teams that track stateful work across multiple runs — contact enrichment, outreach sequences, research queues, content pipelines — a workspace document accumulates state alongside the run thread. Rather than the thread being the record of what happened, the document is the record of where everything stands right now.
Open the document panel from the room header to see the current state of every row: what the agent has filled in, what is still pending, what is blocked, and what has reached a terminal state. The document persists indefinitely — it accumulates across every run, not just the most recent one.
When an agent completes a batch run, it outputs a JSON summary of changed rows that SmartOutput renders inline in the thread as a diff table — so you can see exactly what moved and to what state without opening the document. The panel is there for deeper inspection, not for reading every run result.
Documents support export to CSV or JSON from the room thread at any time: just ask "export the outreach list as CSV." Export does not affect the document's state.