AlertINT / Incident investigation01
AlertINT

Infrastructure incidents, decoded.

AlertINT investigates Zabbix and Prometheus alerts as they fire, so the why is ready before anyone opens a laptop.

Try it on your alerts →

Set up and run for you by an SRE with 10+ years on call.

Fig. 1 — Repeated checkout, payment and frontend alerts on the on-call phone. Illustrative: the services are invented.

While you scrolled, AlertINT already started investigating.

While you scrolled, AlertINT already started investigating.

Example incident: checkout purchases are blocked. AlertINT checks deploys, logs and metrics and finds an expired payment client certificate. The engineer reads the finding through MCP, renews the certificate and records a note. AlertINT confirms recovery. Use the buttons to explore each step.

AlertINT is checking what changed.

#incidents
AlertINTAPP+1m 30s
Investigating
✓ Recovery detected All 12 alerts clear

Correlated 12 alerts on checkout, payment, frontend, ordersPayment's client certificate expired, not the checkout deploy

Alerts:

    📝 Operator note: Renewing payment's client cert now.

      Real investigation sample

      Checkout couldn’t reach payment. The errors showed up next door.Lab run · 7 alerts · 5 services · See the investigation2m 12sto the cause
      Fig. 2 — An example incident, played step by step. The services, times and cause are invented.

      Built for production access.

      1. Read-only by design

        Investigates production without changing it.

      2. Runs in your network

        Deployed alongside your services. It sends no telemetry.

      3. Your LLM

        The provider you choose, or a self-hosted model so prompts never leave your network.

      4. Tamper-evident audit log

        Every investigation step and operator note, hash-chained and verifiable.

      Open source, shipping since June 2026, now on v0.14.x. Read the code ↗

      Works with

      • Prometheus / Alertmanager
      • Zabbix
      • Loki
      • Sentry
      • Slack
      • ntfy
      • Claude Code
      • Codex

      Not seeing your integration? I’ll add it. ↗

      AI model leaderboard / Incident response

      Which AI would you put on call?

      Compare Haiku, Sonnet, Qwen and Gemma on real OpenTelemetry Demo faults. Explore answer accuracy, confidence, speed and cost in the AlertINT AI Leaderboard.

      Explore the AI leaderboard →

      Try it on your own alerts.

      We set AlertINT up on your most critical services, keep it running in your network, and review the results with you. One fixed fee.

      Try it on your alerts →

      Ernests Knavins, founder. 10+ years SRE/DevOps on-call · LinkedIn ↗