CoreModels

CoreModels is schema intelligence middleware. It is the place where you write down what your business objects actually are — the types, the fields, the taxonomies, and the constrained relationships between them — and keep that definition somewhere every other system can read.

It is not a database, not a search engine, not an API platform, and not a data pipeline. It sits alongside those and gives them something to agree with. Most estates already hold this knowledge; it is just spread across three schemas, a wiki page, and one senior engineer's memory. CoreModels gives it one home, with a shape, a history, and an export button.

CoreModels is developed by ARAMAI. Ariesnet configures, deploys and operates it as the authoring layer of the Context Supply Chain — the same product our Implementation and Managed Operations engagements install and run.

• Data and platform teams whose systems each define Customer slightly differently, and who would rather map that than migrate it
• API and integration engineers who need a JSON Schema, an XSD and a JSON-LD context to stay in step with one another
• AI teams who want their agents reading a governed model instead of inferring structure from prose
• Anyone who has been asked in a meeting which system is authoritative for a term, and did not have a document to point at

You model with Types, Elements, Taxonomies and Relation Groups. Relation Groups carry From and To constraints, so the model is genuinely constrained rather than merely drawn — a relation that should only run from an Order to a Customer cannot quietly run somewhere else.

Projects and Spaces let one model span systems. Where two systems mean the same thing by different names, you say so once: either as a mapping relation between spaces, which stays live as the model changes, or as a mixin holding the external system's own identifier, or both together. The choice is yours, and none of it requires the other system to be renamed.

Work happens in the open. Five roles cover the usual spread from read-only reviewer to owner, every change lands in a change log with who, what and when, and comments attach to the exact node under discussion with @mentions to pull in the person who knows. Slack and Trello integrations carry that conversation to wherever your team already is.

Here is the whole list, with no roadmap items smuggled in. If a format is not on this table, it does not ship today.

GraphQL and OpenAPI are available as live API channels rather than as downloadable files. That distinction matters when you are planning an integration, so we draw it rather than rounding it up.

This is our favorite feature, and it is a real one rather than a slogan. The same model renders as Types, Elements and Taxonomies for a data modeler, or as Objects, Properties and Association Types for an API designer. Same model, same constraints, different words on screen.

So nobody has to learn your modeling team's dialect in order to use the model, and no team has to give up the words it thinks in. That is the whole point: agreement about meaning does not require agreement about naming.

CoreModels is core to Stage 2, Structure, and Stage 3, Semantics, and it is strong at Stage 4, Validation, because it authors and exports the conformance contract that other stages enforce. It is not the retrieval engine and it is not the agent runtime — Stages 5, 6 and 7 are served by other things.

We would rather you knew that on day one than discovered it in month three. If your problem is that retrieval keeps returning plausible but wrong material, or that your agents have nowhere governed to write back what they learn, CoreModels is a good part of that answer and not the whole of it. Ask us and we will tell you which part is missing.

• The in-product AI assistant suggests; it cannot modify your schemas. It will answer questions about the model, find types by description, explain a schema in plain language and propose structure from a file you attach — but a person or your own API makes every change.
• It is not the runtime gateway. It authors and exports the conformance contract; enforcing that contract at every boundary is a separate job.
• It does not do governed write-back or usage telemetry. Those are Stages 6 and 7, and they live elsewhere.
• It is not an ingestion pipeline. It imports schemas and structured files; it does not crawl your estate.

Personal and Team both come with a 14-day free trial, and a card is required to start it. Add-on packs stack on Personal and Team when a project count or node count gets tight; Enterprise is unlimited, so the packs do not apply.

One policy worth stating plainly, because it is unusual and it is good: reaching a plan limit does not block work you have already done. Existing projects stay open and editable. You are offered an add-on, an upgrade, or the option to carry on as you are — and that is a choice, not a wall.

CoreModels does the authoring work in the middle of the context supply chain. It is where structure gets typed and where meaning gets agreed across systems, and it exports the conformance contract that later stages enforce at their boundaries.

It is not the retrieval engine, and it is not the agent runtime. Getting the right context to an agent at the right moment, governing what agents write back, and closing the learning loop are handled elsewhere in the chain. We would rather you knew that now than found it out in week three.

See how the whole chain fits together.