Incident Responder
Calm at 3am: incident triage with ranked hypotheses, blameless postmortems, and follow-the-steps runbooks.
by Kenji Yoshida5.0(1)693 installs
About this skill
Use case
Incident tooling (Sentry and friends) ranks high on every MCP list. Triage bias: stop the bleeding first — severity call, blast radius, and the single fastest check per hypothesis. Postmortems come out blameless; runbooks assume a sleepy on-call.
Example prompts
- "Triage: p99 latency 8x on checkout, deploys frozen since Friday"
- "Draft the postmortem from this incident timeline"
Connect in under a minute
- Choose a plan (the free tier needs no card).
- Copy your license key from the dashboard.
- Add the endpoint to your MCP client:
claude mcp add incident-responder https://stg.agently.tokyocorp.net/mcp/incident-responder \
--transport http \
--header "Authorization: Bearer YOUR_LICENSE_KEY"
Works the same in Cursor and ChatGPT developer mode — every successful tool call is metered, failed calls are never billed.
Tools included
triage_incidentTriage a production incident: severity, hypotheses, next diagnostic stepspostmortem_draftDraft a blameless postmortem from an incident timelinerunbook_generatorGenerate an operational runbook for an alert or procedureReviews
5.0(1)The runbooks changed our on-callDataForge Labs · 7/20/2026
Every alert now links to a generated runbook we reviewed once. New joiners handle pages solo in week one.