What You Can Do with MCP

Copy-and-paste prompts for looking things up, investigating incidents, running an incident and reviewing how you respond.

Once your assistant is connected, you don't call tools by name. You just ask. Copy any prompt below, replace the [brackets] with your own details, and paste it into your assistant.

Good to know: Prompts that change things (declaring an incident, posting an update, assigning a role, changing severity, paging someone) affect real incidents and real people, the same as doing it in FireHydrant. Your assistant should ask you to confirm before it acts on the higher-impact ones. Read what it says it's about to do before you approve.

Look things up

See what's happening right now

List my active incidents, newest first, with their severity and current milestone.

Uses incident search.

See what's firing

What Signals alerts are open right now? Show the newest first and tell me which ones are linked to an incident.

Uses alert search.

Find out who is on call

Who is on call right now for the [team name] team, and which services does that team own?

Uses team lookup.

Find someone's teams

Which teams is [person's name] a member of?

Uses user and team lookup.

See what you can choose from when declaring an incident

What severities, priorities, incident types and milestones can I use in our organization?

Uses incident settings.

See your incident roles

What incident roles do we use, and what does each role do?

Uses incident roles.

Investigate

Get the full picture of one incident

Give me the full picture of incident [number or name]: what happened, who is involved, its severity and milestone, and where it stands now.

Uses incident details and timeline.

Summarize a timeline

Read the timeline for incident [number or name] and summarize the key decisions, who made them and when.

Uses the incident timeline.

Summarize the meeting bridge

Summarize what was said on the meeting bridge for incident [number or name]: the main points, decisions and open questions.

Uses the incident transcript. This only works if the meeting was recorded or transcribed, for example by FireHydrant's scribe bot. If there's no transcript, the assistant should say so.

Find related incidents

Find incidents about [topic, for example "Slack" or "voice calls"] and tell me what they have in common: causes, severities and how long they stayed open.

Uses incident search and details.

Check who responded to an alert

For alert [alert ID or name], who was notified and how, and did anyone acknowledge it?

Uses alert details.

Start a retrospective

Using the details and timeline of incident [number or name], draft a retrospective outline: a short timeline, impact, contributing factors and suggested follow-ups.

Uses incident details and timeline.

Run an incident

Declare an incident

Declare an incident named "[short description of what's wrong]" at [severity]. Tell me exactly what you'll set before you create it.

Uses incident settings, then creates the incident. This runs your incident runbooks, which usually open a Slack channel and can notify responders.

Assign a role

Assign [person's name] as [role, for example Commander] on incident [number or name].

Uses user lookup, incident roles and role assignment. The person gets a Slack message and is invited to the incident channel.

Post an update

Post this update to incident [number or name]: "[your update]".

Posts to the incident timeline and its Slack channel.

Move an incident to a new milestone

Move incident [number or name] to [milestone, for example Mitigated] and post this update: "[your update]".

Moving to a resolved or closed milestone can start your retrospective and stop runbooks.

Add a note

Add this note to incident [number or name]: "[your note]".

Adds an internal note that everyone on the incident can see. It never goes to a status page.

Change severity or other details

Change the severity of incident [number or name] to [severity].

Uses edit incident. Changing severity, priority or type can start runbooks that notify people.

Page someone

Page [person's name]: "[what's wrong and why you're paging them]". Use [HIGH, MEDIUM or LOW] priority.

Sends a real notification to their phone or devices and creates an alert.

Review and improve

Look for trends

Review my last 20 incidents and tell me about any trends: recurring causes, the mix of severities, how many are still open, and anything that looks inconsistent.

Uses incident search and details.

Find stale incidents

Which of my active incidents look stale? For each, tell me how long it has been open, its milestone, and whether you'd recommend closing it or looking into it.

Uses incident search.

Find missing or mismatched data

Find incidents that have no severity set, or where the severity and priority don't match, and list them so I can clean them up.

Uses incident search and details.

Look at one area over time

Show me incidents about [service or topic] from the last [3 months]. Is there a pattern, and has it gotten better or worse?

Uses incident search and details.

Tips for better results

  • Be specific. Use incident numbers, team names and people's names. The more precise the prompt, the fewer follow-up questions.
  • Ask to see the plan first. For anything that changes things, add "tell me what you'll do before you do it".
  • Chain steps. For example, "find my stale incidents, then draft a closing note for each one for me to review."
  • Check the result. Assistants can misread data. Open the incident in FireHydrant before you act on a summary.

If there's something you want to do that isn't here, let us know.


Did this page help you?