Prometheus

Prometheus is the standard metrics collection and alerting system in the cloud-native ecosystem — adopted as a CNCF graduated project and deeply integrated into Kubernetes, the Prometheus data model has become an industry standard. Its HTTP API enables agents to query time-series metrics using PromQL, evaluate alert rules, and inspect the health of monitored targets. For agents performing performance analysis, anomaly detection, or automated remediation, the Prometheus query API provides direct access to the metrics layer.

The Prometheus HTTP API covers instant queries (PromQL expression at a point in time), range queries (PromQL over a time range), series metadata, label values, and target health. Agents can execute PromQL expressions to retrieve CPU utilization, error rates, request latencies, and any custom metrics exposed by applications. The Alertmanager API (companion service) provides programmatic access to active alerts and silence management.

Reliability for self-hosted Prometheus is the team's operational responsibility. Prometheus uses local disk storage by default — high-availability deployments use Thanos or Cortex for distributed, durable metrics storage. The scrape model means metric availability depends on both Prometheus uptime and target availability. Teams running Prometheus in production should implement appropriate retention, storage, and HA configuration.

Authentication in stock Prometheus is basic — the scrape API is unauthenticated by default in many deployments. Production deployments add reverse proxy authentication (TLS + BasicAuth) or use the built-in TLS and BasicAuth configuration. The Alertmanager API similarly requires network-level or proxy-level authentication in most deployments.

Documentation is thorough and authoritative — the official Prometheus docs cover PromQL, the data model, configuration, and the HTTP API with strong reference quality. The PromQL documentation is the most important reference for teams building metric query automation. Teams evaluating Prometheus versus Grafana Mimir or VictoriaMetrics for metrics infrastructure should compare the HA and long-term storage approaches.