We built a self-hosted OpsGenie replacement with AI root cause analysis — and it runs completely air-gapped

Hey DevOps community 👋
Like many of you, we got the Atlassian email. OpsGenie is shutting down April 2027.
We looked at the alternatives. PagerDuty and Better Stack are SaaS — hard no for regulated environments. Grafana OnCall just entered archive mode. Everything self-hosted was just a router — it knew who to call but had zero idea why the alert fired.
So we built what we actually needed.
Wachd is a self-hosted alert intelligence platform that runs entirely inside your Kubernetes cluster. One Helm chart. No data leaves your network.
When an alert fires from Grafana, Datadog, or any webhook source — instead of just waking up an engineer with an alert title — it automatically:
Fetches the last git commits to the affected service
Pulls recent error logs and metric history
Correlates the timeline — alert at T, last deploy at T-6min, error spike at T+1
Strips all PII before any AI touches it
Sends the on-call engineer the probable cause in plain English
The AI runs locally via Ollama — Llama 3, Mistral, or Phi-3 on your own hardware. Zero external API calls. For teams comfortable with cloud AI, one config line switches to Claude or OpenAI.
It also includes full on-call scheduling, rotation management, escalation chains, override calendars, AD and SSO integration, SMS and voice via Twilio, and a Slack bot with action buttons.
Apache 2.0. Everything open.
We built it for the teams the SaaS vendors cannot serve — banks, fintechs, healthcare, defence — anywhere data residency is non-negotiable.
Would love feedback from anyone running on-call in regulated environments or evaluating OpsGenie replacements.
👉 github.com/wachd/wachd
🌐 wachd.io
https://www.producthunt.com/products/wachd