# Getting started

From sign-up to your first load, in the eight steps of a working Amfelt setup.

## Getting started — From an empty account to your first load.

These eight steps are for teams with private-preview access. Each links to the guides that cover it in depth.

- **Create your organization**: Sign up and create the organization that owns your projects, members, and billing. Invite teammates as admin, operator, or viewer.
- **Create a project**: A project scopes your sources, data platform, and agents. Its configuration repository is created automatically with the project, so there is no repository step. Your project configuration lives in Git, and every change is a version your team can review. Amfelt creates and hosts it for the project, and you can connect your own Git repository instead. The setup runs in the project's Development environment.
- **Connect an agent**: An agent is a small program in your network that moves the data. Install it on a host with the install script, a package manager, or a release archive; container and Kubernetes installs are on the roadmap. It connects outbound, registers with your organization, and runs your project once you assign it.
- **Connect your data platform**: The data platform is the database where Amfelt writes your loaded data. Saving the connection tests it. Credentials go to the vault (the encrypted secrets store Amfelt keeps separately for each organization), never into config files.
- **Connect a data source**: A data source is a system Amfelt reads your data from. Point the agent at it and test the connection. Once the connection passes, Amfelt reads the source's tables.
- **Choose your data**: Select the tables to load. Each load is a full load, which copies every record. You can set a repeating schedule for the job afterwards.
- **Run the first load**: Start the load. The agent writes the tables into your data platform. The setup shows the load as running until it finishes, with a link to the run in Monitor.
- **Monitor your runs**: Follow execution as it happens, see where a failed run stopped, and read the metrics across runs.