Open-source observability platform with ML-powered Sift investigations and an AI assistant that generates PromQL/LogQL queries from natural language. Adaptive Telemetry automatically drops high-cardinality data before indexing, cutting ingest costs. The open-core model lets you self-host Grafana OSS free or use managed Cloud tiers.
Grafana Labs develops an open-core observability platform comprising four main components: Loki (logs), Tempo (distributed traces), Mimir (long-term metrics), and the Grafana visualization layer. All four are built on object storage backends (S3, GCS, Azure Blob) and designed around a query-time cost model rather than full ingest indexing.
Deployment options: self-host the full LGTM stack under AGPLv3, use Grafana Cloud (managed SaaS), or deploy Grafana Enterprise on-premises with commercial support. Data portability is maintained across all three paths.
Sift correlates logs, metrics, and traces during active incidents to surface probable root cause signals. The AI query assistant generates working PromQL and LogQL from natural-language descriptions. Adaptive Telemetry drops high-cardinality dimensions and low-signal metrics in the ingest pipeline before storage, reducing costs without manual cardinality management. Grafana Alerting supports multi-dimensional rules that fan out across every matching label combination.
Key Features
LGTM stack: Loki for logs, Tempo for distributed traces, Mimir for long-term metrics storage — all built on object storage (S3/GCS/Azure Blob), queried rather than fully indexed, with cost proportional to query volume not ingest volume
AI-generated PromQL and LogQL: natural-language prompt to working query in the editor — reduces the barrier for engineers who read dashboards but don't write metric queries daily
Sift AI incident correlation: automatically correlates logs, metrics, and traces during an active incident to surface probable root cause and related signals across the stack
Adaptive Telemetry: drops high-cardinality dimensions and low-signal metrics in the ingest pipeline before storage — a configurable cost-reduction lever rather than manual cardinality management
Open-source core (AGPLv3): self-host the full LGTM stack, run on Grafana Cloud, or deploy Grafana Enterprise on-premises; data portability is maintained across all deployment models
Grafana Alerting with multi-dimensional rules: one alert rule fans out across every matching label combination, reducing the number of rules to manage across large fleets of homogeneous services
Integrations
12 total
monitoring
DatadogOpenTelemetryPrometheus
cloud
AzureAWSGCP
incident
PagerDuty
ci / cd
Terraform
scm
GitLabGitHub
messaging
Slack
orchestration
Kubernetes
Pricing
3 tiers
Free
Free
Free tier: 10k active metrics, 50GB logs/mo, 50GB traces/mo, 14-day retention, 3 active AI users
