prometheus-metrics-analyzer

name	prometheus-metrics-analyzer
description	Analyze metrics to debug alerts or incidents

Metrics Analyzer Skill

Prometheus MCP Tools

Tool	Purpose
mcp__prometheus__prometheus_list_metrics	List all available metrics
mcp__prometheus__prometheus_metric_metadata	Get metadata (type, description) for a metric
mcp__prometheus__prometheus_list_labels	List label names for filtering
mcp__prometheus__prometheus_label_values	Get values for a label (e.g. service names)
mcp__prometheus__prometheus_query	Run an instant PromQL query
mcp__prometheus__prometheus_query_range	Run a range query (use for last 30 mins)

Debugging Workflow

1. Find relevant metrics — Use prometheus_list_metrics to browse available metrics, then prometheus_metric_metadata to confirm type and description. Select up to 5 metrics most relevant to the issue. You don't have to always select 5 metrics if fewer metrics are enough to debug the issue

2. Query each metric for the last 30 mins — Use prometheus_query_range with start = now-30m, end = now, step = 60s.

3. Find error patterns — Look for spikes, drops, elevated rates, or flatlines. For counters, query the rate. For histograms, query the p99 latency or error ratio.

4. Explain the issue — Summarize what each metric shows, correlate patterns across metrics, and state the likely root cause.

Rules

• Query at most 5 metrics per investigation.
• Always use a 30-minute range unless the user specifies otherwise.
• Prefer metrics with error, failure, timeout, latency, or saturati in the name when debugging incidents.
• Use prometheus_label_values to scope queries to a specific service or namespace if known.