Data Sources

The security boundary that holds your tables, files, and custom tables.

A data source is a container for data in Wallabi — and, more importantly, a security boundary. Everything inside it is accessed, queried, and shared together. When you grant someone access to a data source, you’re granting access to everything it holds; when you don’t, none of it is visible to them.

What a Data Source Holds

A single data source can hold any mix of:

  • External tables — tables from a connected database or application, such as Snowflake, Postgres, or HubSpot
  • Files — uploaded spreadsheets and data files, such as CSV, XLS, or Parquet
  • Custom tables — tables native to Wallabi that you create and manage yourself

Whatever their origin, the tables in a data source look the same to the rest of Wallabi. They appear side by side in the Data Explorer, and a transform can join across them in a single query — an external tickets table against a custom priority_mappings table, for example.

A Security Boundary

The data source is the unit Wallabi uses to decide who can see what:

  • Access is granted per data source. A new data source is private to you until you share it with your organization. Once shared, everyone you’ve shared it with can query every table inside it.
  • Queries respect the boundary. Your apps and Claude Code can only query the data sources you have access to. Credentials for external systems are held by Wallabi at the data source level and never appear in your app code.
  • Relationships live inside the boundary. Foreign keys between custom tables connect tables within the same data source.

Because of this, it’s worth deciding up front what belongs together. Tables that the same people need, and that you’ll want to join, belong in the same data source. Data with a different audience — finance figures that only a few people should see, say — belongs in a data source of its own.

Creating and Sharing

Data sources are created and managed in the Wallabi app. See Connecting to Data for the step-by-step setup, including how to share a data source with your organization.

You can also start with an empty data source to hold only custom tables — ask Claude Code to create one when you need somewhere new to put reference data and no existing data source fits.