Hey PH 👋 I'm the maker behind Nightshift Labs. On-call has two soul-crushing moments: the 2am page where you burn the first five minutes just figuring out what's on fire, and the postmortem nobody wants to write afterward. I built the Incident-Ops Kit to take the edge off both. It's two n8n workflows + Claude. When a PagerDuty alert fires, you get a structured triage in Slack within seconds — what's on fire in plain English, likely blast radius, what changed recently, suggested first moves, and a confidence level. When the incident resolves, it drafts a blameless postmortem: timeline, contributing factors written as questions for the team, action items. The thing I obsessed over: it doesn't bluff. If the data's thin it says "insufficient signal" instead of inventing a root cause — anyone who's carried a pager can smell a confident-but-wrong tool instantly. It runs entirely on your own n8n and your own Anthropic key (cents per incident) — I host nothing and never see your data. There's a free single-workflow version to try first, too. Would genuinely love feedback from anyone who's been on-call. What would make you actually trust an AI triage bot?

Incident-Ops Kit

AI on-call triage + auto postmortems (n8n + Claude)

AI on-call triage + auto postmortems (n8n + Claude)