How it works
A check is a SQL query that runs on a schedule and evaluates its result against a condition. When the condition is breached, Alertee opens an incident in your Inbox and notifies your team — via email, Slack, or webhook. You get a check two ways, and you choose per check:- Write it yourself. Plain SQL plus a condition. It stays yours — Alertee never edits it.
- Let Alertee’s agent author it. Connect a database (or describe what you want in plain English) and the agent inspects your schema, drafts a baseline-aware query, picks a threshold, and keeps it tuned on a review cadence. Checks the agent owns are marked managed by Alertee. See Managed checks.
Beyond the database
Data quality is the core, but a check doesn’t have to be SQL. Point one at a URL or API endpoint to catch downtime, slow responses, or a bad status code, or have a scheduled job send a heartbeat so Alertee alerts the moment that job stops checking in. And if credentials can’t leave your network at all, run the query yourself and push us the number — we alert on it exactly the same way. Same conditions, same Inbox, same alerts.Key concepts
Checks
A check is a SQL query + a condition + a schedule. It’s the core building block.
Managed checks
Let Alertee’s agent author a baseline-aware check from your schema and keep it tuned.
Inbox
Failed checks become incidents you acknowledge, assign, resolve, and classify.
Connections
An encrypted, read-only link to a PostgreSQL or ClickHouse database.
Alert groups
Group your notification channels (email, Slack, webhook) and assign them to checks.
CLI & MCP
Drive Alertee from your terminal or an AI agent over MCP.