Self-hosted · Fair Source · Read-only by design

Infrastructure alerts,
decoded.

Monitoring tools tell you what fired, but rarely enough to investigate. AlertINT is a self-hosted agent runtime that turns infrastructure alerts into incident context your AI agent reads directly over MCP.

alertint — serve

$ alertint serve

12:19:19INFOalertint startingversion=dev

12:19:19INFOrules loadedbaseline · 3 rules

12:19:19INFOprometheus connectedprometheus:9090

12:19:19INFOlogs connectedloki:3100

12:19:19INFOmcp listening:9912/mcp

12:19:19INFOwebhook listening:9911

12:19:19INFOintegration health: okprometheus

12:21:06INFOwebhook receivedalerts=1 · EndpointDown firing

12:21:06INFOcorrelator: new incidentgroup=cluster=prod

How it works

From alert to resolution

When an alert fires, AlertINT gathers the context around it — correlated alerts, metrics, and logs — then posts a finding to Slack. Your AI agent connects over MCP to inspect live incident state and resolve it.

01 alerts fire 02 context gathered 03 finding posted 04 investigate over MCP Alertmanager Zabbix (or others) Prometheus Loki AlertINT self-hosted runtime · MCP server firing resolved resolved Slack finding + next steps Operator + AI agent any MCP client
Alert firing
Runtime path — automatic
Operator path — via MCP

Read-only

Observes and reports by default, never touches your infrastructure.

Self-hosted

Your alert data and incident context stay inside your infrastructure. We enjoy privacy as much as you do.

Fair Source

The core software is Fair Source and always will be.

MCP-first

Your AI agent inspects real incident state and live data, not pasted text.

Get it

A single self-hosted binary.

Available on GitHub. Run the server, fire the built-in incident drill to watch it produce an AI finding end to end, then point your monitoring tools at it and connect MCP.

$ alertint serve --config config.yaml

Starts the AlertINT runtime — ingress, correlation, triage, MCP server.

$ alertint drill --config config.yaml

Runs a synthetic incident through the real pipeline to an AI finding. See quickstart.