# Getting started

> Start with a coding agent or from the command line.

Choose the path you want to use. Both create the same plain YAML and SQL project.

## With a coding agent

DRE is built to be driven by coding agents as well as people: everything is plain YAML and SQL files, every command but `dre init` runs without prompts, and `--log-format json`, `dre validate --json`, `dre ls --output json` and the [manifest](https://getdre.com/docs/manifest/) give tools machine-readable output.

DRE’s [agent skills](https://github.com/get-dre/dre/blob/v0.1.2/skills/README.md) turn Claude Code, Codex, Cursor, GitHub Copilot, Gemini CLI or any agent that reads Agent Skills into a guide, from installing DRE to a delivered report. They ask one question at a time with a recommendation, follow [DRE’s practices](https://getdre.com/docs/practices/), confirm before anything is delivered, and never ask for your passwords or tokens.

```bash
claude plugin marketplace add get-dre/dre && claude plugin install dre@dre   # Claude Codenpx skills add get-dre/dre#skills-latest                                     # other agents
```

Then ask: “help me with dre”.

## From the command line

1.  [Install DRE](https://getdre.com/docs/install/) for your platform.
    
2.  Create a starter project:
    
    ```bash
    dre initcd my_reports
    ```
    
3.  Check that the project works before editing it:
    
    ```bash
    dre validate
    ```
    

A successful validation means the project configuration and SQL compile. Continue with [Build and run reports](https://getdre.com/docs/building-reports/) to add reports, preview output and deliver files.

[Edit page](https://github.com/get-dre/dre/edit/master/docs/getting-started.md)

[Next  
Install](https://getdre.com/docs/install/)