Create provisioning configuration files under /etc/grafana/provisioning/datasources/ and /etc/grafana/provisioning/alerting/ in YAML format
Define each datasource with a unique name, type, url, access mode, and any secureJsonData fields for credentials
For contact points, define the contactPoints section under the alerting provisioning folder with a name, type (email, slack, pagerduty), and settings block
Set deleteMissingDatasources: true on the datasources file to remove stale datasources that are no longer in the provisioned config
Mount the provisioning directory into the Grafana container at startup; changes require a Grafana restart (or hot reload for alerting rules)
Known gotchas
Datasources provisioned via files cannot be edited in the Grafana UI; the UI shows them as read-only to prevent config drift
Secret fields (passwords, API keys) provisioned via plain YAML are visible in the filesystem; use environment variable interpolation ($ENV_VAR) instead
Alert contact point provisioning overwrites all contact points matching the name on every restart; managing partial updates requires careful file structure
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