Skip to main content

​
What Are Runs?

A run is a single execution of your agent’s workflow. Every time your agent performs its task—whether triggered manually, by schedule, or by event—that’s a run.

Manual Run

You click “Test Run” button

Scheduled Run

Runs at 9 AM every day

Event Run

Triggered by new email or webhook

API Run

Executed via API call

​
Run Statuses

Every run has a status that tells you what’s happening:
What it means: The agent is currently executingWhat you see:
  • Real-time progress in the run log
  • AI reasoning step-by-step
  • Updates as data is processed
Actions:
  • Watch live progress
  • Cancel if needed
What it means: The agent finished successfullyWhat you see:
  • Full execution log
  • All outputs written
  • Summary of what was done
Indicators:
  • Green checkmark icon
  • “Completed” badge
  • Final timestamp
Actions:
  • Review results
  • Download logs
  • Re-run if needed
What it means: The agent encountered an errorWhat you see:
  • Error message
  • Stack trace (if available)
  • Where it failed in the workflow
Indicators:
  • Red X icon
  • “Failed” badge
  • Error details in logs
Actions:
  • Read error message
  • Debug the issue
  • Fix and retry
What it means: The agent is waiting for human approval (human-in-the-loop)What you see:
  • What the agent wants to do
  • Why it needs approval
  • Options to approve/reject/modify
Indicators:
  • Yellow pause icon
  • “Needs Review” badge
  • Notification sent (if configured)
Actions:
  • Review the decision
  • Approve to continue
  • Reject to stop
  • Modify and approve
What it means: You or another user stopped the runWhat you see:
  • Partial results (what was done before cancellation)
  • “Canceled by [user]” message
  • Timestamp of cancellation
Indicators:
  • Gray ban icon
  • “Canceled” badge
Actions:
  • Review partial work
  • Start a new run if needed

​
Viewing Run History

Access all past runs from the Runs tab.
1

Navigate to Runs

In your agent page, click the Runs tab in the left sidebar
2

Browse History

See a list of all executions sorted by date (newest first)
3

Filter & Search

Use filters to find specific runs:
  • Status (Completed, Failed, etc.)
  • Date range
  • Trigger type
4

Click for Details

Click any run to see full logs and details

​
Run List View

Each run in the list shows:
ColumnDescription
StatusVisual indicator (✅❌⏸️)
Start TimeWhen the run began
DurationHow long it took (or “In progress…”)
TriggerManual, Schedule, Event, API
Items ProcessedHow many rows/items (if applicable)
UserWho triggered it (for manual runs)

​
Understanding Run Logs

Logs show you exactly what your agent did and why.

​
Log Structure

Chronological view of all actions:
[09:00:01] Run started (Manual trigger by [email protected])
[09:00:02] Reading instructions...
[09:00:03] Loading worksheet "Quote Requests"
[09:00:05] Found 10 rows with Status = "Pending"
[09:00:07] Processing row 1: Acme Corp
[09:00:09]   - Searching Price List for Product A
[09:00:10]   - Found price: $50.00
[09:00:11]   - Calculating total: 10 × $50 = $500
[09:00:12]   - Applying 10% discount: $500 - $50 = $450
[09:00:13]   - Writing to row 1 of Quotes sheet
[09:00:14] Processing row 2: TechCo
...
[09:02:30] All items processed successfully
[09:02:31] Run completed (Duration: 2m 30s)

​
Log Details

Click any log entry to expand and see:
  • Full context: What led to this action
  • Input data: What the agent received
  • Output data: What the agent produced
  • Timing: How long this step took
  • Citations: Which sources were used

​
Debugging Failed Runs

When a run fails, follow these steps to identify and fix the issue.

​
Debugging Workflow

1

Read the Error Message

The error message tells you what went wrong:Common errors:
  • “Source not found” → Source was deleted or renamed
  • “Column ‘XYZ’ not found” → Column name changed or misspelled
  • “API rate limit exceeded” → Too many requests to external service
  • “Formula error in cell B5” → Worksheet formula is broken
2

Find Where It Failed

Look at the execution timeline to see the last successful step before failure.Example:
[09:00:10] Processing row 5: StartupXYZ ✓
[09:00:12] Processing row 6: GlobalCorp ✓
[09:00:14] Processing row 7: FailCo ❌
            Error: Product "Widget Z" not found in Price List
Row 7 is the problem.
3

Reproduce in AI Assistant

Use the AI Assistant to test the failing scenario:
Process row 7 only and show me what happens
This lets you debug interactively.
4

Identify Root Cause

Common causes:Data issues:
  • Missing required fields
  • Invalid formats (wrong date format, non-numeric text)
  • Unexpected values
Configuration issues:
  • Source deleted or renamed
  • Integration disconnected
  • Column names changed
Logic issues:
  • Instructions don’t handle edge case
  • Missing validation rules
  • Incorrect formula in worksheet
5

Fix the Issue

Based on root cause:Data issues: Clean the data, add validation Configuration issues: Reconnect integration, re-upload source Logic issues: Update instructions, add error handling
6

Retry the Run

Click Retry on the failed run to re-execute with the fix

​
Common Failures & Solutions

Error: Source "Price List 2024.xlsx" not foundCause: Source was deleted, renamed, or never uploadedSolution:
  1. Go to Sources tab
  2. Check if source exists
  3. If renamed, update instructions with new name
  4. If deleted, re-upload the source
Error: Column "Customer Name" not found in worksheetCause: Column was renamed, deleted, or misspelled in instructionsSolution:
  1. Open the worksheet
  2. Verify column name (check spelling and case)
  3. Update instructions to match exact column name
  4. Or rename column to match instructions
Error: Gmail integration failed: Authentication expiredCause: Integration was disconnected or credentials expiredSolution:
  1. Go to Settings → Integrations
  2. Find the failing integration
  3. Click Reconnect
  4. Authorize again
  5. Retry the run
Error: Formula error in cell E5: #DIV/0!Cause: Worksheet formula dividing by zero or referencing invalid cellsSolution:
  1. Open worksheet and navigate to cell E5
  2. Check the formula
  3. Fix the formula (e.g., add IF check for zero)
  4. Save worksheet
  5. Retry run
Error: Run timed out after 10 minutesCause: Processing too much data or very slow external APISolution:
  1. Process in smaller batches (e.g., 100 rows at a time)
  2. Optimize sources (remove unnecessary documents)
  3. Simplify instructions (fewer complex steps)
  4. Check external APIs (slow response times)

​
Run Actions

​
Manual Run

Click Test Run to execute your agent immediately. Use when:
  • Testing new agent
  • Processing one-off task
  • Debugging an issue
  • Verifying a fix

​
Retry Run

Click Retry on a completed or failed run to execute again with the same inputs. Use when:
  • Run failed due to temporary issue (API downtime)
  • You fixed the problem and want to reprocess
  • Testing if your fix worked
Note: Retry uses current instructions, not the instructions from the original run.

​
Cancel Run

Click Cancel while a run is in progress to stop execution. What happens:
  • Agent stops processing
  • Partial results may be saved
  • Run marked as “Canceled”
Use when:
  • Realized you need to make a change
  • Processing wrong data
  • Emergency stop needed
Canceling is immediate and cannot be undone. Some changes (like sent emails) cannot be reversed.

​
Download Logs

Click Download Logs to export the full execution log as a text file. Use when:
  • Sharing with support team
  • Archiving for compliance
  • Detailed offline analysis
  • Debugging complex issues

​
Performance Monitoring

Track your agent’s performance over time.

​
Key Metrics

Success Rate

Formula: (Completed runs / Total runs) × 100%Good: > 95% Needs attention: < 90%Improve by:
  • Better error handling
  • Data validation
  • More robust instructions

Average Duration

What to track: Median run timeBenchmarks:
  • Simple tasks: < 30 seconds
  • Medium complexity: 1-3 minutes
  • Complex workflows: 5-10 minutes
Improve by:
  • Reduce source count
  • Optimize worksheet structure
  • Process in parallel

Items Per Run

What to track: How many items processed per runBenchmarks:
  • Small batches: 10-50 items
  • Medium batches: 50-500 items
  • Large batches: 500+ items
Optimize by:
  • Batch processing
  • Parallel execution
  • Efficient data structure

Manual Review Rate

Formula: (Runs needing review / Total runs) × 100%Good: < 10% (for most workflows) Too high: > 30% (agent isn’t confident)Improve by:
  • Better training data
  • Clearer instructions
  • More comprehensive sources
Track success rate weekly or monthly:
Week 1: 85% (testing phase)
Week 2: 92% (refined instructions)
Week 3: 97% (stable)
Week 4: 98% (optimized)
Goal: Upward trend toward 95%+

​
Advanced Run Management

​
Scheduled Runs

View upcoming and past scheduled executions. Upcoming runs:
  • Next execution time
  • Trigger configuration
  • Expected duration (based on history)
Past scheduled runs:
  • Filter by “Scheduled” trigger type
  • See execution pattern (daily, weekly, etc.)
  • Identify missed runs (if any)

​
Bulk Actions

Manage multiple runs at once:
  • Retry all failed runs from the past 24 hours
  • Download logs for all runs in a date range
  • Export run data to CSV for analysis

​
Run Notifications

Get alerted when important events happen:
Configure in Settings → Notifications:Options:
  • Run completed (all runs or only failures)
  • Run needs review (human-in-the-loop)
  • Run duration exceeds threshold
  • Run failed after 3 retries
Recipient: Your email or team distribution list
Post to Slack channel when:Options:
  • Agent completes run
  • Agent fails
  • Agent needs approval
  • Daily summary (runs completed, failed, avg duration)
Setup:
  1. Connect Slack integration
  2. Go to Settings → Notifications
  3. Enable “Send to Slack”
  4. Choose channel
Call your API when runs complete:Payload:
{
  "event": "run.completed",
  "run_id": "run_abc123",
  "agent_id": "agent_xyz",
  "status": "completed",
  "duration_ms": 45000,
  "items_processed": 100,
  "timestamp": "2025-01-15T09:30:00Z"
}
Use case: Integrate with your own monitoring/alerting system

​
Best Practices

Monitor regularly: Check run history at least weekly
Investigate failures immediately: Don’t let failed runs accumulate
Track trends: Look for patterns (time of day, specific data causing issues)
Set up notifications: Don’t rely on manually checking
Archive old logs: Download and store logs for compliance/audit
Test before scheduling: Always manually test before enabling schedule
Document fixes: When you debug an issue, note the solution for future reference

​
Troubleshooting

Possible causes:
  1. Filter applied: Clear all filters
  2. Wrong agent: Make sure you’re viewing correct agent
  3. Date range: Expand date range to “All time”
Still missing? Contact support with agent ID and approximate run time.
Possible causes:
  1. Already completing: Agent is finishing up
  2. Permission issue: You need Editor role or higher
  3. Browser issue: Refresh page and try again
Stuck? Runs auto-timeout after 10 minutes if no response.
Possible causes:
  1. Old run: Logs older than 90 days are archived (Enterprise has longer retention)
  2. Log level: Some details hidden based on plan
  3. Incomplete run: Run was canceled before logging completed
Need archived logs? Contact support to retrieve.

​
Next Steps

Working with Agents

Learn how to optimize agent instructions

AI Assistant

Debug runs interactively with chat

Integrations

Connect services for notifications and actions

API Reference

Programmatic access to run data