Connections and credentials

Database connections overview

A database connection is a saved, named pointer to an external database that Kenseme can read from. An organization admin sets it up once, and everyone in the organization can then pick it by name wherever a feature needs to reach that database.

In the app: Settings › Organization Settings › External Credentials › Database Connections

[screenshot]
The Database Connections page, with connections grouped by database system and the Add Connection button.

Before you start

  • You need the Admin role in the organization to open the Database Connections page. Other members can use saved connections but can’t add, edit, or delete them. See Roles in Kenseme.

The Database Connections page

Open Organization Settings, go to the External Credentials tab, and click Database Connections.

The page groups connections by database system, with a count for each group. Every row shows:

Column What it shows
Name The name you gave the connection.
Server or Target The server the connection points at. For Fabric connections the column reads Target and shows the Fabric item and its workspace.
Description Your optional note.
In Use How many binding environments currently use the connection.
Updated When the connection was last changed.

Each row has Edit and Test buttons and a delete button. The Add Connection button at the top adds a new one.

What a connection is used for

You create a connection once, and these features can all use it:

  • Schema import. Import a schema straight from a live database instead of pasting DDL or uploading a file. See Create a schema.
  • Ontology generation. The ontology generator can draft an ontology from the tables and columns it reads through a connection. See Generate an ontology with AI.
  • Binding an ontology to a database. A binding maps your ontology’s classes and attributes onto real tables and columns, so the Data Agent and the ontology’s MCP server can answer questions against live data. See Bind an ontology to a database.

Note: Only the third item depends on the Bind to Database feature being on for your organization. Creating connections, importing a schema from one, and generating an ontology from one work on every plan.

Supported database systems

You can save connections for six systems:

  • SQL Server
  • PostgreSQL
  • MySQL
  • Databricks
  • Fabric Warehouse
  • Fabric Lakehouse

The first four use a server address and credentials that you type in. The two Fabric systems work differently: you browse your Microsoft Fabric tenant and pick the warehouse or lakehouse, and no connection string or password is stored. They need a Fabric credential to be set up first.

What Kenseme reads

To build a schema, draft an ontology, or capture a binding snapshot, Kenseme reads catalog metadata only: schemas, tables, views, columns, their types, and declared keys. It doesn’t copy your row data for any of these.

The Data Agent and the ontology MCP server are different: they run read-only SELECT queries and return rows to the person asking. That’s why a read-only database login is the right choice for every connection. It limits what a saved credential could ever do.

Tip: Name each connection after what it points at and which environment it serves, such as Sales Warehouse (Prod). You pick connections by name later, when you create a binding environment.