Go Report Card GitHub actions CI Docker Pulls GitHub binaries download

Dynamically generate Grafana dashboard based on Prometheus metrics

Have you ever needed to debug issues and ended up querying Prometheus for group({app="foo"}) by (__name__) to find out what metrics the app exposes and than querying all of them fo find anything suspicious? Or do you often encounter apps that do not have any official dashboard?

Well I have a good news for you, Autograf have you covered! 🎉

How does it work?

Autograf generates Grafana dashboard directly form /metrics or based on a metrics matching provided selector. Each metric has own panel optimized for its type and those are grouped based on metric namespacing. If you want Autograf can event upload the dashboard Right your to a Grafana for you!



Using prebuilt binaries, Docker image of build it yourself.

go install[email protected]


make build

How to use?

./autograf --help
Usage: autograf

Autograf generates Grafana dashboard form Prometheus metrics either read form a /metrics endpoint or queried form live Prometheus instance. The dashboard JSON is by default printed to stdout. But can also upload
the dashboard directly to your Grafana instance. You can configure most of the flags using config file. See the docs.

Example from /metrics:

    curl | autograf --metrics-file -

Example from Prometheus query:

    GRAFANA_TOKEN=xxx autograf --prometheus-url --selector {app='foo'} --grafana-url

  -h, --help                                       Show context-sensitive help.
      --debug                                      Enable debug logging
      --metrics-file=STRING                        File containing the metrics exposed by app (will read stdin if se to - )
      --open-metrics-format                        Metrics data are in the application/openmetrics-text format.
      --prometheus-url=STRING                      URL of Prometheus instance to fetch the metrics from.
      --selector=STRING                            Selector to filter metrics from the Prometheus instance.
      --grafana-variables=GRAFANA-VARIABLES,...    Labels used as a variables for filtering in dashboard
      --grafana-url=STRING                         URL of Grafana to upload the dashboard to, if not specified, dashboard JSON is printed to stdout
      --grafana-folder=STRING                      Name of target Grafana folder
      --grafana-dashboard-name=STRING              Name of the Grafana dashboard
      --grafana-data-source=STRING                 Name of the Grafana datasource to use
      --open-browser                               Open the Grafana dashboard automatically in browser

Loading data from all metrics exposed by app

curl -q | ./autograf --metrics-file -

Loading data from live Prometheus instance

Print Grafana dashboard JSON for all metrics matching selector {instance="", job="prometheus"} form the configured Prometheus instance.

autograf --prometheus-url --selector '{instance="", job="prometheus"}'

Uploading dashboard directly to Grafana

GRAFANA_TOKEN="xxx" autograf --prometheus-url --selector '{instance="", job="prometheus"}' --grafana-url --grafana-folder test
Dashboard successfully generated, see

Config file

If you do not want to set all the flags again and again you can use a config file. By default autograf looks for it in ~/.autograf.json but can be changed using the AUTOGRAF_CONFIG env variable.

Config file syntax

    "prometheus_url": "",
    "grafana_url": "",
    "grafana_dashboard_name": "Autograf",
    "grafana_folder": "FUSAKLAS garbage",
    "grafana_datasource": "Prometheus",
    "grafana_token": "xxx",
    "open_browser": true

Than you can use simply just this!

autograf --selector {app='foo'}

Future ideas

  • Could be a Grafana app plugin and user could just go to{foo="bar"} and the dashboard would show up right in the Grafana itself.
  • Allow customizing the graph visualization using some tags in metric HELP(panel type, aggregations, units, colors, description, …)
  • Add custom visuals for well known metrics


View Github