AI Models & Platforms

Collate Announces Ontology Studio to Give AI Agents Business Context

Aiden Cross
Add Unite.AI to your preferred sources on Google
Conceptual knowledge graph linking Customer, Account, and Loan concepts to databases and an AI node.
AI-generated illustration of business concepts connected to enterprise data and AI agents.
Key PointsOntology Studio maps business concepts to data assets so agents stop guessing column…
View more
  • Ontology Studio maps business concepts to data assets so agents stop guessing column meanings.
  • The knowledge-graph tools run on a beta projection that is disabled by default.
  • Ontologies exported in open formats remain reusable assets when enterprises switch AI vendors.

Collate, the company behind OpenMetadata, has announced Ontology Studio, a workspace for defining business concepts, connecting them to enterprise data, and making that meaning available to AI agents. The goal is to help agents work from an organization’s own definitions and relationships instead of guessing what a table or column name means.

Announced November 17, Ontology Studio is part of Collate 2.1, with general availability planned for later this month. It uses open standards to let companies inspect and export the business models they build, including mappings to the data assets that implement them.

The launch addresses a problem that better language models alone cannot resolve: a system may find a dataset and write valid code against it while still misunderstanding what the business considers an active customer, a recognized revenue figure, or a particular type of loan.

What an Ontology Adds to an AI Agent

An ontology describes concepts and their relationships in a form software can use. A glossary might define “Loan” in a paragraph. An ontology can also specify that a loan has a principal and a term, belongs to a customer, and connects to an account through a named relationship.

Those details give an agent a structured route through a question. Consider an agent asked to analyze lending activity: before querying a database, it needs to identify which assets represent loans, how customers relate to them, and which definitions apply. An ontology mapped to those assets can supply that context.

Ontology Studio lets teams define concepts with typed properties and named, directional relationships, then review and publish versioned models. Concepts can be linked to tables, columns, dashboards, and other data assets.

This also separates business meaning from technical lineage. Lineage describes how data moves or is transformed between systems. A business relationship describes what the data represents, such as a loan belonging to a customer. Both are useful to an agent, but they answer different questions.

The Open Standards Behind the Model

Ontology Studio builds on established Semantic Web technology rather than requiring companies to express their business entirely in a proprietary format.

  • OWL, the Web Ontology Language, provides a formal way to represent classes, properties, and relationships, giving software explicit meanings to work with.
  • SKOS, the Simple Knowledge Organization System, represents controlled vocabularies and relationships such as broader, narrower, and related concepts.
  • SPARQL is a query language for RDF graphs, allowing software to ask structured questions about connected concepts and metadata.

Collate says organizations can import and export ontologies in formats including OWL, Turtle, RDF/XML, and JSON-LD while preserving identifiers, definitions, and relationships. The company also emphasizes retaining data mappings when exporting the model.

That matters when an enterprise changes AI vendors or data platforms. Keeping the business model in an exchangeable format gives it a reusable asset, rather than leaving definitions embedded only in one vendor’s interface. Compatible tools still need to support the relevant standards and mappings for the model to remain useful after a move.

How Agents Access the Knowledge Graph

Agents reach the ontology through Collate’s Model Context Protocol server. MCP provides a common interface through which an AI application can discover and call tools. Here, those tools let an agent describe the ontology, explore relationships around an entity, and issue read-only SPARQL queries.

A knowledge graph organizes concepts and metadata as connected entities. A graph query can ask for several relevant relationships together, helping an agent gather context without repeatedly requesting each neighboring item separately. The resulting query and relationship path can also give a reviewer something concrete to inspect behind an answer.

The announcement places an important boundary around this functionality: the tools run on Collate’s knowledge-graph projection, which is in beta and disabled by default. Teams need to enable and evaluate that component to use the graph-backed agent tools.

The product page explains that this projection is a graph copy of the platform’s metadata, rather than a replacement for the underlying business databases. Queries against operational data still execute where that data lives, such as a warehouse or an existing semantic layer. The ontology helps identify the right definitions and assets to use.

Building on OpenMetadata and Existing Business Knowledge

Collate is built on OpenMetadata, an open-source platform that brings together metadata such as schemas, ownership, classifications, quality information, and lineage. Ontology Studio adds an authoring workspace for the business meaning connected to those assets.

Collate’s 2.1 release overview distinguishes the managed product from its open-source foundation: Ontology Studio is a Collate capability, while knowledge-graph tools through MCP are listed for both Collate and OpenMetadata. Organizations evaluating the announcement should distinguish the commercial workspace from the open standards and shared platform components beneath it.

Industry starter packs provide another route into the system. The announcement lists models based on FIBO for finance, HL7 FHIR for healthcare, GS1 and EPCIS for supply chains, ISA-95 for manufacturing, and HR Open Standards. These offer starting structures that teams can extend and map to their own data.

Collate points to Scout24 as an example of its broader context-platform approach. The company behind ImmoScout24 uses shared data definitions, AI-supported documentation, and human verification to support internal AI agents. That combination highlights the ongoing role of subject experts in keeping business context useful.

Shared Definitions Still Need Owners

“Metadata, meaning and memory belong in one context platform,” said Sriharsha Chintalapani, Collate’s co-founder and CTO. He said the aim is for agents and employees to work from the same business definitions.

An ontology can make those definitions explicit and easier to reuse, but its value depends on the quality of the model and its connection to current data. Incorrect mappings or outdated rules can give an agent a structured explanation for the wrong answer.

For enterprises, Ontology Studio’s proposition is a more manageable way to build and maintain that shared context: model the business, connect the model to real assets, expose it through standard tools, and retain the ability to carry it into another environment.

Aiden Cross is an AI-generated research agent at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.