Read and manage agent runs
Every agent run is recorded with its plan, its summary, the records it touched and its credit cost. You can open a run while it is still going, stop it, and ask the agent follow-up questions afterwards.
Nothing an agent does is hidden. Each run keeps the plan it wrote for itself, the evidence behind each step, the records it created or updated, and the full conversation. Reading one run properly is usually faster than rewriting an agent's instructions on a hunch.
Where runs show up
- The agent dashboard. Click any agent row. Recent runs is at the bottom of the page.
- The sidebar. Expand the Background Agents section, hover an agent, and a popover shows its five most recent runs with their statuses, clickable straight into the run.
- Chat. Ask the assistant what an agent did, for example "what did my lead gen agent do yesterday?".
An agent with no runs yet says "This agent's first run will appear here the next time its trigger conditions are met."
What the agent dashboard tells you
Four headline metrics sit at the top: Runs, Next run, Credits used and Last activity. Below them:
| Panel | What it holds |
|---|---|
| Setup answers | What you told the builder during setup, with a link back to the setup conversation |
| Daily spend | A credit spend chart with a period selector |
| Records | The Up next and Already run tabs, with eligibility and next opportunity per record |
| Recent runs | The run list |
| Agent setup | Schedule, Cost, Run scope, Created |
| Capabilities | Which of Email, Slack and Enrichment are enabled |
What a run page shows
Open a run and you get, in order:
- The run plan. The checklist the agent wrote for itself, each step marked
Pending,In progress,Done,BlockedorCanceled, carrying the evidence behind it. - An executive summary. What the run concluded, in a paragraph.
- Entities modified. The CRM records the run created or updated, grouped by type, with counts of new versus updated and links into your CRM.
- The activity flow. Step by step, what the agent read, searched and wrote.
- The full conversation. Everything the agent said to itself while working.
The header carries the status, when it ran, the trigger type and the credit cost. While a run is still going the page shows what it has finished so far plus Still working, and it refreshes itself while the tab is open.
Stopping a run
Open the active run
Go to Background Agents, open the agent, and click the run in Recent runs. Active runs are marked with a spinner.
Click Stop process
It is in the top right. Confirm in "Stop this process?", or click Keep running to leave it alone.
Read what survived
The run flips to
Canceled. Steps that were already done stayDonewith their evidence; steps that were pending or in progress now readCanceled. Partial output and diagnostics stay readable.Check for staged changes
If the agent had staged CRM changes before stopping, they are still in Proposals waiting for you.
Only active runs can be stopped. A finished run returns "Only active runs can be stopped".
Asking about a run afterwards
At the bottom of a finished run is an Ask a follow-up question... box. The conversation continues from the run's own thread, so the agent still has the full context of what it did and why. Use it for the questions the summary does not answer: why it skipped a company, where a number came from, what it would need to do better next time.
What runs cost
Every run is metered in credits. The price per run is shown in the Cost column on the agent list, in the Agent setup panel on the dashboard, and in the confirmation dialog before any manual run. Some agents charge a fixed price per run, others charge per unit they deliver, and some are free.
Below the agent list, Usage History itemises every charge with Type, Agent, Record, Charged to, Cost and Date. Failed attempts appear there too, labelled "Out of credits" or "Daily limit exceeded". The Analytics tab charts the same data.
For what credits cost and how your allowance works, see Credits and usage or talk to us.
Limits
- Only the 10 most recent runs per agent are kept in the run list, and the sidebar popover shows 5.
- A run has a wall-clock ceiling of 12 hours. A run still marked active after that is displayed as
Canceled, because a stalled run never reaches a final status on its own. - Run history refreshes only while the browser tab is visible.
If something goes wrong
A run has said "Running" for hours and never finishes. It stalled without reaching a final status. Open it and click Stop process, or wait: after 12 hours it is shown as Canceled automatically. Completed steps and their evidence stay readable either way.
"Could not load runs" or "Could not load this run". The run history is temporarily unavailable, usually during a deploy. Click Try again. Nothing is lost: runs that already completed are still recorded.
"Run not detected. The agent run did not appear." The run was accepted but did not surface within 30 seconds. Reload the agent dashboard. If it truly never starts, check the credit balance and whether the agent is paused.
A run page says "No messages in this run". The run produced no conversation, usually because it failed very early. Check the header for the status, then look in Usage History for a matching failed row labelled "Out of credits" or "Daily limit exceeded".
Every agent is paused and admins were emailed. The workspace ran out of credits mid-run. Top up in Settings > Billing, then resume each agent with the play button. More symptoms in Agent problems.