Skip to main content
Every time an agent runs, Slant logs the execution. Use run history to understand what agents are doing, verify results, and troubleshoot issues.

View run history

1

Open the agent

Navigate to AutomationsAgents and select an agent.
2

Click Runs

Click the Runs tab to see execution history.
3

Browse runs

View recent runs with their status and timing.

Run statuses

StatusMeaning
CompletedAgent finished successfully
Pending reviewAgent completed but actions await approval (draft mode)
RunningAgent is currently executing
FailedAn error stopped the agent
CancelledRun was manually stopped

Run details

Click any run to see details:

Trigger information

  • What triggered the run — Event, schedule, or manual
  • When it started — Timestamp of execution
  • Record context — Which record(s) the agent worked with

Reasoning trace

See how the agent analyzed the situation:
  • Input data — What information the agent received
  • Analysis — How the agent interpreted the context
  • Decisions — What actions the agent chose and why
This transparency helps you understand agent behavior.

Actions taken

Review what the agent did:
  • Tasks created — Links to any new tasks
  • Notes added — Content appended to records
  • Fields updated — Changes made to record data
  • Notifications sent — Alerts delivered to team members
Each action shows the before and after state when applicable.

Output

For agents that produce content (summaries, drafts):
  • View the generated text
  • See how variables were resolved
  • Check the quality of the output

Review pending actions

In draft mode, agents queue actions for your approval:
1

Open pending runs

Navigate to the agent’s Runs tab and filter for Pending review.
2

Review the proposed actions

Click a run to see what the agent wants to do.
3

Approve or reject

  • Approve — Execute the proposed actions
  • Reject — Discard without executing
  • Edit and approve — Modify before executing
Note: Pending actions expire after 7 days if not reviewed.

Troubleshoot failed runs

When an agent fails, investigate:
1

Open the failed run

Click on the run with Failed status.
2

Check the error message

Read the error description to understand what went wrong.
3

Review the reasoning trace

See where in the process the failure occurred.
4

Fix the issue

Adjust configuration or data as needed.
5

Retry if appropriate

Click Retry to re-run with the same inputs.

Common failure causes

ErrorCauseFix
”Record not found”The record was deletedNo action needed
”Permission denied”Agent lacks required accessCheck agent permissions
”Rate limit exceeded”Too many runs too quicklyWill retry automatically
”Invalid configuration”Settings issueReview agent configuration
”Context too large”Too much input dataNarrow agent scope

Filter and search runs

Find specific runs using filters:
  • Status — Completed, Failed, Pending, etc.
  • Date range — Runs within a time period
  • Trigger type — Event, schedule, or manual
  • Record — Runs for a specific household or contact

Export run data

Download run history for analysis:
1

Apply filters

Set filters to select the runs you want.
2

Click Export

Click Export to download as CSV.
3

Review in spreadsheet

Analyze patterns, timing, or outcomes.

Run metrics

View aggregate statistics:
  • Total runs — How many times the agent has run
  • Success rate — Percentage of successful completions
  • Average duration — How long runs typically take
  • Actions per run — Average number of actions taken
Use metrics to evaluate agent effectiveness and identify issues.

Cancel a running agent

To stop an agent mid-execution:
1

Find the active run

Look for runs with Running status.
2

Click Cancel

Click Cancel to stop execution.
3

Confirm

Confirm cancellation. Actions already completed cannot be undone.
Note: Cancellation is best-effort. Some actions may complete before the cancellation takes effect.

Best practices

Review runs regularly. Check agent activity weekly to ensure agents behave as expected. Investigate failures promptly. Persistent failures often indicate configuration issues. Use draft mode initially. Review agent decisions before trusting automatic execution. Monitor success rates. Declining success rates may indicate changing data patterns.

Next steps