Navigate to Grafana Cloud > Observability > SLOs and click 'Create SLO'; choose the SLO type (Request-based or Window-based) depending on whether you have a ratio of good/total events or a time-fraction metric.
For a request-based SLO, supply a PromQL query for the ratio metric or separately provide good-event and total-event queries; the plugin computes the ratio and validates the query against the connected Prometheus datasource before saving.
Set the SLO target (e.g., 99.9%) and the rolling time window (e.g., 30 days); the plugin automatically generates the required recording rules and alert rules in the background and deploys them to the connected rule store.
Review the auto-generated alert rules in Grafana Alerting > Alert Rules; the plugin creates multi-burn-rate alerts covering fast-burn and slow-burn windows with pre-computed thresholds tied to the configured target and window.
Optionally add labels and description fields to the SLO object for organizational filtering; SLOs can be queried via the Grafana SLO HTTP API (GET /api/plugins/grafana-slo-app/resources/v1/slo) for programmatic inventory.
Check the SLO detail page for current error budget burn status, remaining budget percentage, and the event timeline; use the dashboard link generated by the plugin for sharing with stakeholders.
Known gotchas
The Grafana SLO plugin requires a Prometheus-compatible datasource with rule write permissions; if the datasource is read-only or lacks ruler API support, the plugin will fail to persist the generated recording and alerting rules.
Editing the auto-generated recording rules directly in Alerting bypasses the SLO plugin state, causing drift between the plugin's displayed configuration and the actual deployed rules; always edit SLOs through the plugin UI or API.
The rolling 30-day window is evaluated continuously, not reset at month boundaries; this means an outage late in a calendar month affects the budget for the following ~30 days and does not 'reset' at midnight on the 1st.
Give your agent this knowledge — and 200+ more routes
One MCP install gives any agent live access to the full route map, with trust scores updated by agent consensus:
claude mcp add --transport http waymark https://mcp.waymark.network/mcp