Set up your DAC project
Install DAC and scaffold a starter project with sample dashboards, a semantic model, and an AI agent skill - all from one command.
What you'll do
Install the DAC CLI, run dac init, and walk the project structure it generates. By the end of this step you'll have a my-dashboards/ directory with two starter dashboards, a semantic model, and an AI skill ready to go.
Why this matters
DAC is opinionated about layout - dashboards live in dashboards/, reusable metrics in semantic/, connections in .bruin.yml. Knowing where things live up front makes everything else faster.
Instructions
1. Install DAC
Run the installer. It downloads the dac binary and, if you don't already have it, installs the Bruin CLI too (DAC reuses Bruin connections).
curl -fsSL https://raw.githubusercontent.com/bruin-data/dac/main/install.sh | bash
Verify the install:
dac --version
Tip
Want the bleeding edge? Append -s -- --channel edge to the curl command.
2. Scaffold a project
dac init my-dashboards
cd my-dashboards
The default --template starter gives you a YAML dashboard, a TSX semantic dashboard, a semantic model, and a sample DuckDB-backed connection.
3. Walk the layout
ls -1
cat .bruin.yml
ls dashboards/
You should see:
.bruin.yml- environments and connections (DuckDB by default)dashboards/- one YAML and one TSX starter dashboardsemantic/- a starter semantic model with metrics and dimensionsdata/dac-demo.duckdb- an empty DuckDB file ready to seed.claude/skills/and.codex/skills/- thecreate-dashboardskill that lets Claude or Codex author dashboards for you
What just happened
You have a complete, validated, version-controlled DAC project. The starter dashboards work out of the box because they use inline VALUES blocks for their data - in the next step you'll point a dashboard at a real DuckDB table.