# Ontology Studio

Define your business in an open ontology, and give your AI agents the meaning behind your data

[Get Collate Free](/welcome)

![Ontology Studio](/_next/image?url=%2Fimages%2Fontology-studio%2Fhero.webp&w=3840&q=75)

![Business Meaning Agents Can Use](/_next/image?url=%2Fimages%2Fontology-studio%2Ficons%2Fbusiness-meaning-agents-can-use.svg&w=96&q=75)

Business Meaning Agents Can Use

Agents follow your structured concepts and relationships instead of guessing from names

![Concepts With Real Structure](/_next/image?url=%2Fimages%2Fontology-studio%2Ficons%2Fconcepts-with-real-structure.svg&w=96&q=75)

Concepts With Real Structure

Each concept has its own properties, relationships, and rules

![Open and Yours to Keep](/_next/image?url=%2Fimages%2Fontology-studio%2Ficons%2Fstart-from-a-published-standard.svg&w=96&q=75)

Open and Yours to Keep

Built on open standards like OWL and SKOS, so you can inspect, validate, and export it

## Give agents the meaning behind your data

Expose the model to agents and people, with the query behind every answer

MCP Tools for Agents

One graph call replaces a chain of tool calls, so agents take fewer steps and use fewer tokens

Answers You Can Check

Agents return the SPARQL and the relationships behind an answer, so a person can verify it

SPARQL Console

Write SPARQL with saved queries, shared templates, and RDFS or OWL inference levels

Visual Query Builder

Build a query without writing SPARQL, then open the generated query to read and edit it

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fagent-meaning.webp&w=3840&q=75)

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fagent-meaning.webp&w=3840&q=75)

## Model your business concepts

Give every concept the properties and relationships that define it

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fconcept-classes.webp&w=3840&q=75)

Properties for Each Concept

An Account carries an account ID and a Loan carries a term, each attached to its own concept

Named Relationships

Part of, broader than, calculated from, or your own types, each with a defined meaning

Rules and Constraints

Say what a concept is, what counts as the same thing, and what it can never be

Delete Impact Preview

See what a deletion would break before it happens, such as for child concepts, assets, and mappings

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fconcept-classes.webp&w=3840&q=75)

## See which data carries each concept

Bind the business model to the underlying data and find the gaps

Model and Data Views

Switch from the concept model to the tables and dashboards that implement each concept

Meaning vs. Lineage

See which links come from your business model and which come from data pipelines

Data Binding

See the underlying assets behind each concept, so an unused concept stands out

Ontology Health

Find concepts with no links or no data, and fix each gap from the same screen

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fdata-coverage.webp&w=3840&q=75)

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fdata-coverage.webp&w=3840&q=75)

## Start from a standard. Take it anywhere.

Begin with an industry ontology, extend it for your business, and export it whenever you want

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fstandards-library.webp&w=3840&q=75)

Ontology Starter Library

Pre-packaged ontologies to get started quickly, including FIBO, HL7 FHIR, GS1, and more

Standards-Based Import

Bring in OWL, Turtle, RDF/XML, or JSON-LD ontologies from your existing work

Export With Mappings Intact

Export your ontology with SHACL validation, while keeping identifiers, definitions, and mappings

Dry Run Before Anything Is Written

Preview every concept, relationship, and mapping an install or import will create

![\[object Object\]](/_next/image?url=%2Fimages%2Fontology-studio%2Fstandards-library.webp&w=3840&q=75)

## Built for modern data & AI practices

Designed for changing needs of data & AI teams

AI-Driven Automation

Improve productivity, enforce governance and reduce costs with AI driven automation

Unified Platform

One platform for all your teams for data discovery, observability and governance

Collaborate Around Data

Accelerate development of data assets with social workspaces and knowledge centers

Get started with Collate today for free

[Get Collate Free](/welcome)

Managed Service for Production Data Teams

[Book a Demo](/contact-sales)

## FAQs

Expand All

What is an ontology?+

An ontology is a model of the things a business cares about, written in a form software can act on. Each concept is a class with its own properties, so a Customer has an ID, a name, and an address. Concepts connect through typed relationships, such as an order belonging to a customer or a metric being calculated from others. Standards like OWL and SKOS define how those models are written and exchanged.

Why do AI agents need an ontology?+

An agent can read tables and columns, but a schema says little about what the data means or which business rules apply. An ontology gives the agent the concepts, properties, and relationships the business uses, so it can follow them instead of guessing from names. When the ontology is bound to the data, known as concept realization, the agent knows which tables represent each concept.

What is Ontology Studio?+

Ontology Studio is the workspace in Collate where you build a business ontology, bind it to your data, and make it available to AI agents. You can view, edit, and query the model in one place. In Collate, it replaces Ontology Explorer, adding authoring, standards-based import and export, and agent access. It is the meaning layer of the Collate context platform, alongside the metadata that describes your data and the memory the platform builds up over time.

How do AI agents use the ontology in Collate?+

Collate exposes the knowledge graph through its MCP server as different graph tools: ontology\_describe, sparql\_query, entity\_neighborhood, find\_by\_tag, and shacl\_validate, which an MCP client such as Claude, Cursor, VS Code, or Goose can call. An agent working through them can return the query and the relationships behind its answer, which makes the answer verifiable.

Can I take my ontology with me?+

Yes. Ontology Studio is built on OWL and SKOS, the W3C standards for ontologies and vocabularies. You can inspect the model, validate it against SHACL shapes, and export it with its identifiers, definitions, relationships, and data mappings, so it works with the platforms and agents you choose.

Which standards, formats, and industry ontologies does it support?+

Ontology Studio imports and exports standard ontology formats such as OWL and JSON-LD, with optional SHACL validation. The Ontology Starter Library adds six packs, including FIBO for financial services and HL7 FHIR for healthcare, plus packs for retail, supply chain, manufacturing, and HR.

How does Ontology Studio keep changes governed?+

Only one person can edit the model at a time, so changes never collide. Deleting a concept first shows what it would break. The ontology carries the same ownership, access controls, and audit trail as the rest of Collate.

Do I need to run a graph database?+

No. Your metadata stays in the MySQL or PostgreSQL database Collate already uses, and that database remains the source of truth. Collate keeps a graph copy of it in sync automatically, and the graph is what powers SPARQL, validation, and the agent tools. It ships with your existing deployment, so there is no separate graph database to set up, license, or migrate to. The graph holds the model, not your rows: queries still run in Snowflake, Databricks, or wherever the data lives.

How does Ontology Studio relate to a semantic layer like Cube, dbt, or AtScale?+

They work together. Ontology Studio is independent of query engines and does not replace a semantic layer or a metrics store. Where those tools define metrics, Ontology Studio models the business concepts those metrics belong to, binds them to the assets that implement them, and exposes that model to agents. Teams run both: the ontology tells an agent which definition applies and where the data lives, and the query still executes in Snowflake, Databricks, or your semantic layer.