GreenKube ships with a CLI with two runtime modes: start runs the collector service and demo launches a self-contained sample environment. Reporting and recommendations are provided by the REST API and the dashboard, not by the CLI.
In a Kubernetes deployment the collector container already runs greenkube start; commands can be run inside the pod via kubectl exec:
# Find the GreenKube pod name
kubectl get pods -n greenkube
# Open a shell in the pod
kubectl exec -it <pod-name> -n greenkube -- bash
greenkube startStart the collector service: connects to the database, runs schema migrations, performs an initial collection, then schedules data collection, compression, dashboard summary refresh, the recommendation lifecycle job, and (when automation is enabled) the inline worker loop and discovery runtime.
greenkube start [OPTIONS]
| Flag | Description |
|---|---|
--last TEXT |
Initial backfill window, e.g. 10min, 2h, 7d, 3w, 1m (month). Only used for the initial run; scheduled runs use the normalized window. |
This is the default runtime for the collector container in the Helm chart (command: ["greenkube"], args: ["start"]).
greenkube demoStart GreenKube in demo mode with pre-populated sample data — no live cluster required. Creates a temporary SQLite database, starts the API server and serves the dashboard.
greenkube demo [OPTIONS]
| Flag | Description |
|---|---|
--port INTEGER |
Port for the API server (default: 8000) |
--days INTEGER |
Number of recent days to generate hourly workload metrics for (default: 30); older metric and carbon-intensity history extends to two years at daily resolution |
--no-browser |
Do not open the browser automatically |
Example:
docker run --rm -p 9000:9000 greenkube/greenkube demo --no-browser --port 9000
greenkube versionPrint the GreenKube version.
greenkube.cli.automation_worker)The production profile starts the plan worker as:
python -m greenkube.cli.automation_worker --queue plans
It should run as a single replica. The worker claims queued pull-request operations through database leases, polls linked provider pull requests, and exposes its authenticated control-loop metrics on GREENKUBE_WORKER_METRICS_PORT (Helm: automation.workerMetricsPort, default 8001). Do not run it manually alongside automation.mode=inline; use the Helm chart so the mode and credentials are configured consistently.
| Flag | Description |
|---|---|
--version |
Print the GreenKube version and exit |
--help |
Show help for any command |
The REST API is started by the separate greenkube-api entry point (the API container in the Helm chart). Reports and recommendations are available through the API and the dashboard; see API Reference and Recommendation lifecycle.