Quick overview
Monitor n8n workflow failures, execution health, and AI agent quality from one workflow. OpenAI analyzes incidents and recommends corrective actions, while Slack and Telegram deliver configurable alerts. Includes scheduled health checks, agent evaluation, alert suppression, and optional incident logging for operational visibility.
How it works
- This workflow combines three reliability-monitoring modules into a single n8n automation.
- Incident Detection:* The Error Trigger receives execution failures from monitored n8n workflows. A manual demonstration trigger is also included for testing.
- AI Incident Analysis:* Error information is normalized and sanitized before being sent to OpenAI. Two AI analysis stages classify the incident, identify possible causes, and recommend non-destructive troubleshooting steps.
- Incident Alerts:* The workflow determines severity and applies a configurable alert cooldown to reduce repeated notifications. Incidents can be delivered to Slack and Telegram. Optional n8n Data Tables record incident details.
- Scheduled Health Monitoring:* Every 15 minutes, the system retrieves recent executions and workflow information through the n8n API. It evaluates failure rates, execution duration, and configured critical-workflow inactivity.
- Health Reporting:* Detected issues generate notifications. The workflow also supports daily health summaries and an optional external heartbeat integration.
Setup
- Import the workflow:* Download the JSON file and import it into your n8n instance using the Import from File option.
- Configure OpenAI:* Create an HTTP Bearer Auth credential using your OpenAI API key. Assign it to the AI Triage Analyst, AI Recovery Planner, and AI Quality Evaluator nodes. The default model is
gpt-4.1-mini.
- Connect Slack:* Create a Slack credential with permission to post messages to your selected channel. Assign it to the three Slack notification nodes.
- Connect Telegram:* Create a Telegram bot using BotFather and obtain the destination chat ID. Configure the Telegram credentials on the three Telegram notification nodes.
- Configure notification settings:* Open the Normalize Incident, Health Detector and Digest, and Normalize Agent Trace Code nodes. Update the notification channels, alert thresholds, and enable or disable optional integrations.
- Configure n8n API access:* Create an n8n API credential and connect it to the Get Recent Executions and Get All Workflows nodes. The API credential must have permission to read execution and workflow information.
- Optional audit logging:* Create the
arc_incidents and arc_agent_evaluations Data Tables with the required columns. Select the correct tables in the audit nodes and enable Data Table logging in the configuration.
- Secure the agent evaluation webhook:* Configure a Webhook Header Auth credential with an
X-ARC-KEY header and a strong secret. Other agents can submit execution traces using this authenticated endpoint.
- Test the workflow:* Run the manual demonstration triggers and verify the OpenAI responses, alert conditions, Slack messages, and Telegram notifications.
- Enable monitoring:* Publish the combined workflow. Open each workflow you want to monitor, navigate to Settings, and select this workflow under Error Workflow.
- Verify scheduled monitoring:* Confirm that the 15-minute health scanner executes correctly and that the n8n API returns execution data.
- Complete all configuration and end-to-end tests before using the system in production.
Requirements
- A compatible n8n instance with Error Trigger, Schedule Trigger, Webhook, Code, HTTP Request, and n8n API nodes.
- An n8n API key with permission to read workflows and executions.
- An OpenAI API key with access to
gpt-4.1-mini for AI-powered analysis.
- A Slack workspace and bot credential for Slack notifications (optional).
- A Telegram bot token and destination chat ID for Telegram notifications (optional).
- n8n Data Tables for persistent incident and evaluation records (optional).
- An external heartbeat monitoring service such as Healthchecks.io (optional).
- An authenticated webhook credential for receiving agent evaluation traces.
- Appropriate permissions to configure an Error Workflow on monitored automations.
- External services may have separate usage charges. Credentials are not included in the downloadable JSON.
Customization
- The workflow is designed to be customized for different n8n environments and operational requirements.
- Alert channels: Enable or disable Slack and Telegram independently and configure destination channels or chat IDs.
- AI model: Change the OpenAI model in the HTTP Request nodes to a compatible model available to your API account.
- Monitoring frequency: Adjust the default 15-minute schedule according to your operational requirements.
- Failure thresholds: Customize failure-rate, execution-duration, and critical-workflow inactivity thresholds.
- Alert suppression: Adjust notification cooldown periods to reduce repeated incident alerts.
- Critical workflows: Specify important workflow IDs and their maximum expected inactivity periods.
- Agent evaluation: Adjust quality scoring and operational thresholds for response time, tool usage, and estimated costs.
- Audit logging: Enable optional Data Tables for incident history and agent evaluation records.
- External monitoring: Configure an optional heartbeat endpoint to help detect missed monitoring runs.
- Manual testing: Use the included demonstration triggers to validate the monitoring branches before production use.
- For high-volume environments, extend execution pagination, persistent alert suppression, and monitoring infrastructure as required.
Additional info
Agent Reliability Center (ARC) — AI-Assisted Monitoring Template
This template combines incident triage, execution health monitoring, and AI agent quality evaluation into one workflow with independent triggers.
Primary use cases:
- Monitoring business-critical n8n automations
- Investigating failed executions
- Detecting increased workflow failure rates
- Reviewing AI agent output and tool usage
- Monitoring reported latency and estimated token costs
- Sending operational alerts to Slack and Telegram
- Maintaining optional incident audit records
Security and operational considerations:
AI-generated diagnoses are recommendations, not verified root-cause findings. The workflow does not automatically retry failed business operations, change credentials, delete records, or perform destructive recovery actions.
Agent evaluation depends on the trace information supplied by the calling application. Token counts and estimated costs are not independently verified against provider billing data.
Notification cooldowns use best-effort workflow state and are not designed to guarantee exactly-once delivery across distributed installations.
Testing status: The workflow JSON has undergone offline structural and syntax validation. Each user must configure credentials and complete live integration testing before relying on it in production.
This is an extensible monitoring starter system rather than a replacement for a dedicated enterprise observability platform.