Skip to content
OKStatusOKStatusDocs

Incident response from your AI assistant

An alert comes in at 2 a.m. Without opening the dashboard, you find out what is happening, tell your customers and send the report, in a few sentences. The AI assistants (MCP) guide stays the reference for setting up the connector and for the full list of tools.

The scenario

Section titled “The scenario”

Discord shows: 🔴 Shop · payment API is down. Your assistant, here Mistral Vibe, is already connected to OKStatus. Each exchange below names the tool the assistant calls; the answers are an example. Tools marked (Max) change something and need the Max or Enterprise plan.

Before you start

Section titled “Before you start”
  • The connector added to your assistant: https://mcp.okstatus.eu/mcp (see AI assistants (MCP)).
  • Reading works on every plan. Opening an incident or sending a report needs Max or Enterprise, with a role other than viewer.

1. Find out what is happening

Section titled “1. Find out what is happening”

You: Is anything down right now?

Assistant · get_status_overview 1 monitor down: Shop · payment API, since 02:07. 11 up, 0 paused. No open incident.

You: Show the latest checks of the payment API, region by region.

Assistant · list_monitor_checks 02:13 fr-dc3 timeout (10 s) · 02:13 de-fra timeout (10 s) · 02:13 nl-ams timeout (10 s) · 02:06 fr-dc3 200, 184 ms. All three regions fail the same way: it isn’t a local network problem.

You: And over the last 30 days, where are we?

Assistant · get_monitor_uptime 99.93 % over 30 days, 31 minutes of downtime in total, 7 of them for the current outage.

2. Tell your customers

Section titled “2. Tell your customers”

You: Open a major incident “Payments unavailable” on the payment API: we are investigating.

Assistant · create_incident (Max) Before I open it: title “Payments unavailable”, severity major, first update “We are investigating payment failures.” Shall I go ahead?

You: Yes.

Assistant Incident opened, status “Investigating”. It shows on the Shop status page, and its subscribers have been notified.

You: Add an update: cause found, the load balancer certificate expired, renewal in progress.

Assistant · add_incident_update (Max) Update posted, status “Identified”.

3. Close and report

Section titled “3. Close and report”

You: It’s fixed, mark the incident resolved.

Assistant · add_incident_update (Max) Incident resolved at 02:41. Duration: 34 minutes.

You: E-mail the October SLA report of the Shop group to management@example.com.

Assistant · send_sla_report (Max) Report sent to management@example.com. It counts against your monthly report quota.

Good habits

Section titled “Good habits”
  • Ask for the per-region detail before concluding: one failing region often means a network problem, not an outage.
  • Pick an assistant hosted in the EU if your data must stay there: what the assistant reads is sent to it. We recommend Mistral Vibe.
  • The assistant deletes nothing and doesn’t create alert channels, because a channel holds a secret. Every change goes into the audit log, under your name and the assistant’s.

Going further

Section titled “Going further”