// product · available now
n8n AI Incident Triage Workflows
Alertmanager & Grafana alerts triaged before they reach chat
Importable n8n workflows that turn Alertmanager and Grafana alerts into triaged chat messages: what's wrong, likely cause, first checks, runbook link.
● instant download · secure checkout by Gumroad
// tested
Tested, not just written
Every release runs its full test suite against real tooling before it ships.
Every workflow is imported into a real n8n 2.39.7, published and executed against mocked LLM, chat, Prometheus, Kubernetes, Loki, PagerDuty and GitHub endpoints
A password, AWS key, JWT, bearer token and connection string planted in an alert never reach the model, and the message still arrives
LLM returns 500, hangs or answers broken JSON: the notification still arrives with a fallback note
An Alertmanager group of 3 alerts produces one message listing 3; a resolved group skips the model entirely
Unauthenticated webhook requests are rejected: no run, no model call, no notification
Dedup, flap suppression, PagerDuty trigger and resolve, and per-team routing are each asserted on what the mocks received
// pricing
One-time purchase, three tiers
Pay once, keep the files. Each tier includes everything in the tier before it.
Starter
On-call teams with Alertmanager or Grafana
$19USD
- Alertmanager and Grafana Alerting triage into Slack, Teams, Discord or a generic webhook
- Daily alert noise digest with tuning advice from the Prometheus API
- Shared settings, LLM router and notifier sub-workflows
- Alertmanager config example, Grafana contact point guide, sample payloads
- 30 days of updates
Pro
SRE and platform teams on Kubernetes
$49USD
- Runbook-aware triage: an allowlisted runbook_url fetched into the prompt
- Kubernetes pod failure explainer with pod status, events, previous logs and read-only RBAC
- Deduplication and flap suppression per fingerprint
- PagerDuty Events v2 escalation with auto-resolve, no model in the paging path
- Postmortem drafts, error workflow, prompt customization guide, 1 year of updates
Studio
Consultants, agencies, multi-team platforms
$99USD
- Loki log context around the alert time
- Per-team routing to channels and PagerDuty services by label
- Weekly on-call handover report
- Prompt evaluation kit: 12 golden alerts and a scoring script
- Agency rollout guide, client-use license, 1 year of updates
// contents
16 workflows
- Alertmanager → AI triage → chatStarter+
- Grafana Alerting → AI triageStarter+
- Daily alert noise digestStarter+
- Settings, LLM router & notifierStarter+
- Runbook-aware triagePro+
- Kubernetes pod failure explainerPro+
- Dedup & flap suppressionPro+
- PagerDuty escalation + auto-resolvePro+
- Postmortem draft + error workflowPro+
- Loki log contextStudio+
- Per-team routingStudio+
- Handover report + prompt eval kitStudio+
// compatibility
Works with your stack
- n8n 2.39.7 self-hosted (node types verified against that release)
- OpenAI · Anthropic · Ollama · any OpenAI-compatible server (vLLM, LiteLLM, OpenRouter)
- Prometheus Alertmanager · Grafana Alerting · Prometheus, Mimir, Thanos
- Slack · Microsoft Teams · Discord · PagerDuty · generic webhook
- Kubernetes · Grafana Loki · GitHub
// faq
Questions
+Does my alert data go to OpenAI?
Only what you choose, and only after redaction. Resolved alerts and PagerDuty escalation never call a model at all. With Ollama or your own OpenAI-compatible server nothing leaves your network, and provider 'none' keeps the notifications without AI.
+What happens when the model is down?
The notification still goes out with the raw alert data and a one-line note. That is a tested case, not a hope: 500s, timeouts and malformed JSON each have their own test.
+Does it work on n8n Cloud?
Yes for anything that only reaches the internet. The Kubernetes explainer and a private Prometheus, Loki or Alertmanager need n8n where it can reach them. UI import re-assigns workflow ids, so you re-select the three sub-workflows; the CLI import keeps them.
+How does this pair with the alert rules pack?
Directly. The Prometheus Alert Rules Pack ships a runbook_url on every alert, which is exactly what the runbook-aware triage workflow reads.
+Can I use it for client work?
Starter and Pro are licensed for your own organization. Client and agency use requires Studio.
+Refunds?
If the workflows don't work with the pinned n8n version and the troubleshooting guide doesn't fix it, contact us within 14 days for a refund.
Wake up to an answer, not a label.
Questions first? Email hello@fractaltechware.com