dashboard serve
What this page covers
Serve one or more dashboard drafts through a lightweight local preview server.
When to open this page
- Use this when you are iterating on one dashboard draft, one draft directory, or one generator command and want a local browser surface without publishing back to Grafana after every change.
Who this page is for
Best for SREs, Grafana operators, and responders working with dashboard inventory, migration, inspection, or screenshots.
Purpose
Serve one or more dashboard drafts through a lightweight local preview server.
When to use
Use this when you are iterating on one dashboard draft, one draft directory, or one generator command and want a local browser surface without publishing back to Grafana after every change.
Key flags
--input: local dashboard file or directory to load into the preview server.--script: external generator command whose stdout emits one dashboard JSON or YAML document, or an array of documents.--script-format: parse--scriptstdout asjsonoryaml.--watch: extra local files or directories to watch for reloads.--no-watch: disable background polling reloads.--open-browser: open the preview URL in your default browser after the server starts.--address,--port: bind address and port for the local preview server.
Notes
- This server is a lightweight draft preview and document-inspection surface. It does not embed a full local Grafana renderer.
--inputand--scriptare mutually exclusive. Use--inputfor local draft files and--scriptwhen an external generator already produces the dashboard payload.- Reload errors stay visible in the preview page so you can keep editing without restarting the server.
Examples
# Purpose: Serve one local draft file.
grafana-util dashboard serve --input ./drafts/cpu-main.json --port 18080 --open-browser
# Purpose: Serve all dashboard drafts in one local directory.
grafana-util dashboard serve --input ./dashboards/raw
# Purpose: Serve one generated dashboard and watch generator inputs for reload.
grafana-util dashboard serve --script 'jsonnet dashboards/cpu.jsonnet' --watch ./dashboards --watch ./lib --port 18080