FireHydrant MCP

FireHydrant's MCP server connects an AI client to the incident management platform's API, so responders can work an incident from the same window where they are already reading code and logs. Rather than switching to a browser mid-page, an engineer can ask the agent to open an incident, attach the right severity and impacted services, pull the timeline of what has happened so far, and post an update for stakeholders. The server covers the objects FireHydrant is built around: incidents with their status, severity and affected services, alerts and signals arriving from monitoring, the service catalog and owning teams, and retrospectives with their event timelines and action items. Read access matters as much as write access, because the common agent task is summarizing: which similar incidents hit this service before, which runbook applies, and what actions came out of the last retrospective that nobody closed. It runs as a local server started through npx and authenticates with a FireHydrant API key, so the agent operates with exactly the permissions granted to that key. It suits on-call engineers, incident commanders and reliability teams who want less context switching during an active incident, and faster write-ups afterward, where the agent drafts from the recorded timeline instead of a human reconstructing events from chat history.