Hear about downed Better Stack monitors right in chat

The agent calls the Better Stack Uptime API — it reads monitors and their state, incidents together with their history, heartbeats, status pages and on-call schedules. Out of that a picture comes together: what is down, pending or paused right now, when the last check happened, what went down overnight and for how long, who is on call at this moment and what customers see on the public status page at the same time.

A monitor state and an incident are different things, and the agent does not conflate them: a paused monitor is not checked and creates no incidents, so it is named separately in the summary rather than dissolving among the healthy ones. A heartbeat works the other way round — the alarm is the absence of a signal, and a quiet background job looks fine until its expected window runs out, so the agent looks at the time of the last check, not only at the colour of the status.

By default the agent does not write: it reads monitors and incidents and reports in chat, and it will not create a monitor, change one or take one off pause without an explicit request from you. Incidents arrive over a webhook — creation, acknowledgement and resolution are seen immediately — but that webhook carries no verified signature, so the agent treats a notification as a reason to re-read the state through the API and drops repeat deliveries by the incident identifier.

The token itself bounds the selection: a team key returns the monitors of one team, and the agent names that boundary out loud rather than passing a part of the account off as the whole. Check and incident history comes back in pages windowed by time, so a long period is assembled in parts with the agent pacing its own requests instead of hitting the limit halfway through a report. If a section is closed off by the token rights or there simply is no data for the period, it says so directly instead of showing an empty table.