Hi Hunters, I'm Om — solo founder, ex-on-call. Built Saneops because I watched a 4-person SRE team slowly stop opening their pagers. The reason wasn't laziness. It was math. When 89% of pages are noise — flapping monitors, duplicate alerts from three different tools for the same incident — you eventually learn to mute. Then the one page that actually mattered sits unopened for an hour while production burns. Saneops sits between your monitoring stack (Grafana, Datadog, PagerDuty, Prometheus) and your humans. Three things it does that current tools don't: 1. Correlates related alerts into a single incident using label-match + semantic similarity. 27 alerts from one outage become 1 incident. 2. Drafts a first-pass root-cause analysis with YOUR LLM key — Anthropic, OpenAI, Gemini, Ollama, whatever. Your alert data never leaves your stack. BYOK isn't marketing fluff — it's the only way I'd trust an AI tool with production telemetry. 3. Self-hosts via Docker. If you're regulated and can't send alert payloads to a SaaS, run Saneops on your own infra. In one beta tenant: 11% of pages were actionable before. 70% after 30 days. Same incident volume — different psychological relationship with the pager. Free 60-day closed beta. 1,000 alerts/month. No credit card. Looking for 10 design partners who'll tell me what's broken about it. Genuinely brutal feedback welcome — especially from anyone who's been on-call long enough to have stories. — Om saneops.in

Saneops

Cut on-call alert noise 80% with AI-drafted RCAs

Cut on-call alert noise 80% with AI-drafted RCAs