Self-Healing Error Triage
Set this workflow as an agent's Error Workflow and every terminal failure is triaged by TypeSafe AI's Jev model: retry now, wait and retry, escalate, or ignore. Transient errors — timeouts, rate limits, upstream 5xx — are re-run automatically with the original payload; bugs, bad credentials, and bad data go to your on-call channel. A retry marker guarantees a workflow is retried at most once before a person is paged, so a permanently broken run can never loop.
Templates never consume AI Builder credits. Customize every node visually after import.
How it works
Catch
The agent's error handler starts this workflow with the failed run's name, error message, retry count, and original input.
Guard, then decide
A Code node stamps a retry marker and a Condition node escalates immediately if this payload was already retried once. Otherwise an AI Decision node classifies the error and says whether it looks transient.
Heal or page
Retry branches re-run the failed workflow via Call Workflow with the stamped payload (after a 10-minute delay for outages). Escalate and uncertain branches post to Slack.
Integrations in this template
Every node is swappable — replace any integration with another without rebuilding the flow.
Webhook Trigger
View integration docs →AI Decision
View integration docs →Call Workflow
View integration docs →Built for
- Agents that depend on flaky third-party APIs
- Overnight batch workflows nobody is watching
- Any workflow where a page at 3 a.m. should mean a real bug
Frequently asked questions
How do I wire it up?
Import the template, then open the agent whose failures it should handle and pick this workflow as its Error Workflow. The webhook trigger is only the entry point — the platform starts the run.
What stops an infinite retry loop?
The retry branches add _errorHandlerRetry: true to the payload they re-run. If that run fails too, the handler sees the marker and escalates instead of retrying. Every failure is retried at most once by this workflow.
Can it change the failing parameters?
Not in this template — Jev decides, it does not generate. If you want a repaired payload, add an AI Prompt node on the retry branch to rewrite the input before Call Workflow.
Related templates
Production Incident Controller
Decide ignore, rollback, restart, page, or investigate before an engineer opens Slack.
View template →Confidence-Gated Triage Queue
Route what the model is sure about; queue only the uncertain cases for a person.
View template →Ship this agent today.
Free plan includes everything this template needs. No credit card required.