Installation

Anfra is a local, git-native CLI — no daemon and no cloud account required to get started.

Create a new project

npx create-anfra-project@latest

Or add Anfra to an existing repo:

npx anfra init

Either command sets up the project folder structure, installs the Anfra CLI, and loads the default analytics agent skills.

The exact install channel (npm / pip / single binary) isn't finalized yet — npm is shown here for illustration.

Guided setup

cd your-project
npm install

Open the project in your AI coding assistant (Claude Code, Cursor, etc.) and run the guided setup command:

/anfra-setup

This detects an existing dbt project and offers to import it, or scaffolds a semantic layer from scratch — connecting a warehouse, scanning the schema, and creating starter datasets and metrics. See Project Structure for what gets generated.

Your first query

Once setup finishes, ask a question in plain English from the same agent session:

> What was MRR by plan last month?

The agent resolves the question against your governed model instead of guessing SQL, and shows what it used — the dataset, metric, and dimensions it resolved against.

No agent? Query the semantic layer directly:

anfra query --metric mrr --by plan --grain month
anfra explore   # interactive REPL to browse and compose metrics

Next steps