Help Center
DocsRuns & Metrics
View agent run history, performance analytics, and troubleshoot issues
Run History
View all past runs for an agent
Navigate to an agent's detail page and click the Runs tab to see all past runs.
| Field | Description |
|---|---|
| Status | Completed, Failed, Running, or Cancelled |
| Summary | AI-generated description of what was found |
| Actions | Count of proposed, approved, rejected, and auto-executed actions |
| Trigger | Schedule, Manual, or Event |
| Duration | How long the run took in seconds |
| Dry Run | Test runs that simulate agent execution without applying any changes. Useful for validating agent behavior before enabling auto-execution. |
Use the Status dropdown above the run list to filter by All, Completed, Failed, or Running.
Click a run to see its full detail, including the AI's reasoning and all proposed actions with their statuses. The Activity tab shows a chronological log of everything the agent did during the run, including tool calls, data lookups, and decisions.
Performance Metrics
Track agent effectiveness over time
The Metricstab on an agent's detail page shows:
| Field | Description |
|---|---|
| Total Runs | Number of times the agent has executed |
| Actions Proposed | Total actions generated across all runs |
| Approval Rate | Percentage of actions approved or auto-executed |
| Average Confidence | Mean confidence score of proposed actions |
| Actions Over Time | 12-week bar chart showing approved/rejected/auto-executed trends |
| Top Action Types | Which tools the agent uses most frequently |
Test Runs (Dry Runs)
Simulate agent execution without applying changes
Click Test Runon an agent's detail page to run the agent in preview mode. Test runs:
- Execute the AI with the agent's full instructions and tools
- Generate proposed actions marked as preview (shown with a cyan indicator)
- Never actually modify any CRM data
- Appear in run history with a "[Test Run]" prefix in the summary
Use test runs to validate instructions before activating an agent or switching to "Go ahead" mode.
Rate limit: 5 test runs per user per hour. The error response includes a retry timer so the UI shows the cooldown.
Paused Runs (Ask the User)
When an agent pauses to ask you a question mid-run
An agent can pause mid-run to ask you a clarifying question — for example, "Should I include the Acme renewal in the dropped-deals list, or treat it as a re-engagement opportunity?". Paused runs show a prominent Agent paused — needs your answer card on the run detail page.
- The card displays the question, the context the agent gathered, and (when available) suggested answers you can click instead of typing.
- Submit your answer and the run resumes from where it paused — no loss of context.
- Click Cancel run to abandon the run instead. Already-proposed actions remain in the queue (review or reject them as usual).
- Paused runs do not consume a fresh slot in the per-agent concurrency limit — they hold the existing slot open until you answer or cancel.
Chained Runs
When one agent triggers another on completion
If the agent has a chain target configured, the child run appears in the run history with Trigger: chain and a parent run link. You can navigate up to the parent or down through the chain to trace a multi-step pipeline.
Chains are capped at 30 hops total and skip themselves silently for dry/test runs. See Custom Agents → Chaining for the full behavior including the single-hop cycle-detection limit.
Audit Log
Full execution audit trail
Every tool execution is recorded in the audit log with: tool name, parameters, result, success/failure, and duration. This provides a complete compliance trail for all agent actions. The audit log is maintained automatically and retained for all runs.
Troubleshooting
Common issues and solutions
| Field | Description |
|---|---|
| Agent not running on schedule | Verify the agent status is "Active" (not paused or draft). Check the cron expression matches your intended schedule. |
| Run shows "Failed" | Click into the run to see the error message. Common causes: AI provider not configured, rate limits exceeded, invalid tool parameters, trial period expired, or monthly AI token limit reached. Check Settings > Usage for your current usage. |
| Too many low-quality actions | Refine the agent instructions to be more specific. Increase the confidence threshold. Reduce the max actions per run. |
| Actions expiring before review | Actions expire after 72 hours. Review the queue more frequently or consider enabling auto-execute for low-risk actions. |
| Scheduled run timing is off | Scheduled agents are checked every 15 minutes, so execution may be slightly delayed from the exact scheduled time. For example, an agent scheduled at 8:00 AM may run anytime between 8:00 and 8:15 AM. |