mcp-incident-intel

MCP server wiring PagerDuty, Dynatrace, and runbook context for AI-driven incident response.

Overview

mcp-incident-intel is a Model Context Protocol server that gives AI agents full incident context in a single, standardized interface. It wires together PagerDuty incident data, Dynatrace observability signals, and operational runbooks so that LLM-powered agents can triage, correlate, and begin remediation without switching between tools.

Built for SRE and platform engineering teams in regulated financial services and manufacturing where mean time to resolution is a compliance metric.

Tools exposed

Tool	Description
get_active_incidents	Fetch open PagerDuty incidents with severity and assignee
get_dynatrace_problems	Fetch open Dynatrace problem cards with root cause
get_service_topology	Get upstream/downstream dependencies for a service
get_slo_status	Fetch SLO burn rate and remaining error budget
get_runbook	Retrieve runbook content for a given alert name
correlate_incident	Cross-reference a PagerDuty incident with Dynatrace problems

Quick start

pip install mcp-incident-intel

export PAGERDUTY_TOKEN=your_pagerduty_token
export DYNATRACE_URL=https://your-env.live.dynatrace.com
export DYNATRACE_TOKEN=your_dynatrace_api_token

mcp-incident-intel

Configuration

Variable	Description	Required
PAGERDUTY_TOKEN	PagerDuty API token	Yes
DYNATRACE_URL	Dynatrace environment URL	Yes
DYNATRACE_TOKEN	Dynatrace API token	Yes
RUNBOOK_DIR	Path to runbook markdown files	No

Contributing

See CONTRIBUTING.md.

License

Apache 2.0 — see LICENSE.