MCP PagerDuty

Overview

The MCP PagerDuty server is an early‑stage integration that brings the power of PagerDuty’s incident management platform directly into Model Context Protocol (MCP) workflows. By exposing PagerDuty APIs through MCP, developers can let AI assistants query, create, and update incidents without writing custom HTTP clients. This eliminates the need for manual API calls or third‑party wrappers, allowing natural language interfaces to orchestrate incident response in real time.

At its core, the server implements a set of tools that mirror key PagerDuty endpoints. The most immediately useful are and , which let an AI assistant list active incidents with optional filters and fetch detailed incident data respectively. A third tool, , provides access to the conversation history attached to an incident. These tools are wrapped in MCP’s declarative syntax, so a client can simply request “list all critical incidents for Service X” and receive structured JSON that the assistant can parse or display.

For developers building AI‑powered operations dashboards, this server offers several tangible advantages. First, it abstracts away authentication and pagination logic; the MCP client handles token management via environment variables. Second, because MCP is designed for resource‑first interaction, the server can evolve to expose additional PagerDuty resources—such as escalation policies, incident templates, or service status dashboards—without breaking existing clients. Third, the integration supports sampling and prompting hooks defined by MCP, enabling assistants to ask clarifying questions (e.g., “Do you want to add a note?”) before executing an action, which improves safety and reduces accidental incident creation.

Real‑world use cases include automated on‑call handovers, where an AI assistant can fetch the list of open incidents for a team and suggest priority actions. Another scenario is incident triage: an assistant can pull the latest notes, summarize the issue, and propose next steps to a responder. In CI/CD pipelines, the server can be invoked by an AI bot to create incidents automatically when a deployment fails, ensuring that alerts are logged and tracked in PagerDuty from the moment they occur.

Because MCP is language‑agnostic, any client—Python, Node.js, or even a web browser—can consume the PagerDuty server. Developers can embed the server in their own MCP ecosystem, combine it with other tools (e.g., Slack, GitHub), and build sophisticated AI workflows that span monitoring, communication, and incident resolution—all while keeping the logic declarative and testable.