AI-powered observability assistant for Grafana that helps you query data, investigate issues, and manage dashboards through natural language.
This repository also configures its own agents. See what ask-o11y-plugin tells them →
Skill Claude CodeCodex
Queries AWS CloudWatch metrics through Grafana CloudWatch datasources — discovering namespaces, metrics, dimensions, and running metric queries. Use when the user asks about AWS resources such as EC2, ECS, RDS, Lambda, or Container Insights, or mentions CloudWatch, AWS, or a specific AWS region.
Skill Claude CodeCodex
Analyzes performance issues — CPU, memory, request latency, throughput, and error rates — and identifies bottlenecks by correlating metrics, logs, and traces. Use when the user asks about performance, slow services, latency, saturation, bottlenecks, or resource constraints — even when they say "why is it slow", "the…
Skill Claude CodeCodex
Analyzes distributed traces in Tempo with TraceQL — errors, latency, and call paths between services. Use when the user asks about traces, spans, request flows, service dependencies, latency across services, or debugging "no traces showing" — even when they say "find slow requests", "what called this", or "show the…
Skill Claude CodeCodex
Creates and updates Grafana dashboards panel by panel through the dashboard MCP tools, with correct panel types, units, thresholds, and template variables. Use when the user asks to build, create, modify, or extend a dashboard — even when they say "add a panel", "make a chart", "dashboard for this service", or…
Skill Claude CodeCodex
Investigates firing alerts and incidents and drives root cause analysis across metrics, logs, and traces, prioritizing the alert's runbook. Use when the user mentions alerts, incidents, firing notifications, outages, or asks to find the root cause of a problem — even when they say "something is down", "why did we get…
Skill Claude CodeCodex
Queries continuous profiling data in Pyroscope to find CPU, memory, and contention hotspots. Use when the user asks what is burning CPU, causing high memory, allocations, goroutine leaks, lock contention, or asks for a flame graph — even when they say "why is memory so high", "profile this service", or "find the…
Skill Claude CodeCodex
Renders PromQL, LogQL, and TraceQL results as interactive graphs, gauges, tables, and log or trace panels directly in the chat, with the right visualization type for the data shape. Use when composing answers whose queries benefit from a visualization — even when the user just asks to "show", "graph", or "chart"…
Skill Claude CodeCodex
Writes efficient PromQL for Prometheus/Mimir metrics and LogQL for Loki logs, including label discovery for unfamiliar clusters. Use when the user asks for queries, graph data, metric analysis, or log searches, or before building panel or alert queries — even when they say "error rate", "p99", "how many errors", "log…
Skill Claude CodeCodex
Diagnose and reduce Prometheus/Mimir metrics cost and cardinality — active-series triage, per-metric and per-label drill-down, histogram 14x multiplier, churn diagnosis, label-strategy audits, and Adaptive Metrics recommendations. Use when investigating a high metrics bill, "too many series", "reduce cardinality"…
Skill Claude CodeCodex
Author k6 load-test scripts and Grafana Synthetic Monitoring checks — HTTP/WebSocket/gRPC/browser scripts, executor selection, thresholds, checks vs expect(), smoke/stress/spike/soak patterns, and SM scripted-check semantics. Use when the user says "load test this API", "stress test", "smoke test", "soak test", "spike…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: