Download and run the VictoriaMetrics single-node binary with '-storageDataPath' and '-retentionPeriod' flags; it exposes a Prometheus-compatible remote_write endpoint at '/api/v1/write' on port 8428
Configure existing Prometheus instances (or vmagent) to remote_write to 'http://<vm_host>:8428/api/v1/write'; VictoriaMetrics accepts the same protobuf format as Prometheus remote_write
Replace Prometheus scraping with vmagent by providing a scrape config YAML (same format as prometheus.yml) and starting vmagent with '-promscrape.config=<path>' and '-remoteWrite.url=http://<vm_host>:8428/api/v1/write'
Query data using the MetricsQL-compatible HTTP API at '/api/v1/query' and '/api/v1/query_range' (same paths as Prometheus); point Grafana's Prometheus data source URL to VictoriaMetrics directly
Monitor VictoriaMetrics internals by scraping its '/metrics' endpoint; key metrics include 'vm_rows_inserted_total', 'vm_cache_size_bytes', and 'vm_rows_ignored_total' for ingestion health
Known gotchas
VictoriaMetrics uses MetricsQL which is a superset of PromQL; most PromQL queries work unchanged, but some advanced PromQL functions behave differently or have extended semantics that can produce different results
VictoriaMetrics deduplicates samples within a configurable deduplication interval; when multiple agents write the same metric, ensure '-dedup.minScrapeInterval' is set to match the scrape interval to avoid data loss
The single-node binary does not support clustering; high-availability deployments require the separate cluster version with vminsert, vmselect, and vmstorage components, which have different configuration flags and endpoints
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