Ontologies

Ontologies overview

An ontology is your organization’s shared business vocabulary: the concepts you work with, how they relate, and the facts that describe them. Where a schema says “there is a dbo.CUST table with a cust_id column”, an ontology says “we have Customers, a Customer places Orders, and every Customer has an email address”. It describes the business in its own words, independent of where the data lives.

In the app: Data Context › Ontologies

[screenshot]
The Ontologies list: search and status filter, the four stat tabs, the Recent cards, and the All ontologies table.

What an ontology is made of

You build an ontology from four kinds of building block:

  • Classes are business concepts: Customer, Order, Product. Classes form an “is a kind of” hierarchy, so a Preferred Customer can be a kind of Customer.
  • Relationships connect one class to another: a Customer placesOrder an Order. Relationship identifiers are always written in lowerCamelCase. See Edit an ontology.
  • Attributes are facts about a class: a Customer has a name and an email. Each attribute has a value type such as text, number, or date.
  • Instances are specific named examples of a class, such as “Canada” as an instance of Country. Use them to pin down fixed reference values.

These are the default words. An organization admin can switch the whole editor to formal OWL terms (Object Property, Data Property, Named Individual) or to Microsoft Fabric terms (Entity, Field, Record). See Ontology settings.

How ontologies fit with the rest of Kenseme

  • Schemas are the physical side: tables, columns, and the SQL that builds them. An ontology is the conceptual side. One concept can map to columns spread across several tables. You can also generate a schema from an ontology.
  • Glossaries define terms in prose for people to read. An ontology adds structure: how concepts connect and which facts each one carries. Glossaries and standards can feed the ontology generator.
  • Bindings map classes, relationships, and attributes to real tables and columns. Binding is what lets the Data Agent answer business questions from live data.

The Ontologies list

The list page has:

  • A Search ontologies... box and a status filter (All Statuses, Published, In Review, Draft).
  • Four stat tabs: total Ontologies, Published, In Review, and Draft. Click one to filter the list to that status.
  • A Recent row with the three most recently modified ontologies.
  • An All ontologies section you can show as a table or a grid. The table lists status, namespace, current version, owner, class and relationship counts, a Bind. count, the modified date, and the last Fabric sync. Click the Bind. number to jump straight to the ontology’s bindings.

Ontologies imported from Microsoft Fabric can carry an extra badge next to the status. Hover it for the full explanation:

  • Out of sync: the Fabric ontology it was imported from has changed since. The copy isn’t updated automatically; re-import it if you want the newer version.
  • Source removed: the Fabric ontology no longer exists. Your copy keeps working.
  • Import incomplete: the Fabric import didn’t finish. Kenseme retries it automatically.

Click any row to open the ontology. Use Add Ontology to start a new one.

The ontology page

Opening an ontology shows its status, namespace, last sync, name, and description. A line under the description says where the ontology came from, for example Created blank, Imported from file catalog.ttl, or Imported from standard FIBO. If an AI generation run for this ontology is in progress or waiting for review, a banner at the top links to it.

The toolbar holds the everyday actions:

Button What it does
Edit Change the ontology’s name, description, namespace, owner, and status.
Diagram Show the ontology as an interactive graph. See Visualize the ontology diagram.
Assistant Open the AI chat drawer. See Chat with the ontology assistant.
Export Download the ontology as Turtle or OWL. See Export an ontology.
More Everything else (listed below).
Delete Delete the ontology. It sits apart on the right so you don’t click it by accident.

The More menu contains:

  • Agent Sources: choose the context the AI assistant’s Tool Assisted mode draws on. See Generate an ontology with AI.
  • Add Descriptions: have AI describe every class, relationship, and attribute. See Generate descriptions for an ontology.
  • Duplicate: copy the ontology. See Duplicate an ontology.
  • Prefixes: manage the short prefixes used in identifiers and exports.
  • Attribute groups: assign your organization’s custom attributes. See Apply attributes on ontology entities.
  • Recent changes: review and undo recent edits.
  • Generate Schema (or, once a schema exists, Update Schema, Unlink Schema, Delete Generated Schema, and Bind Generated Schema). See Generate a schema from an ontology.
  • Versions: snapshots of the ontology. See Ontology versioning.
  • Sync history: past pushes and pulls with Microsoft Fabric.
  • Bind to Lakehouse and Bind to Database: map the ontology to real data. An ontology can be bound to a lakehouse or to a database, not both.
  • MCP Server: expose the ontology to AI assistants. See Expose an ontology over MCP.

Bind to Database and MCP Server appear only when those features are turned on for your organization.

Below the header is a collapsed Attributes panel for custom attribute values on the ontology itself, and then the editing workbench, where you add and change classes, relationships, attributes, and instances. See Edit an ontology.

Note: Members with read-only access see the ontology, the diagram, versions, sync history, and bindings, but not Edit, Assistant, Delete, or the editing entries in More. If your organization is in read-only mode, those controls are shown but disabled.

Delete an ontology

  1. Click Delete.
  2. Type the ontology’s name exactly as shown.
  3. Click Delete Ontology.

The ontology and everything inside it are removed from your workspace.

Where to go next

Start an ontology:

Shape it:

Connect it to data and share it:

Manage it: