Introducing Converge

An ontology-centred platform where data, applications and AI share one governed model.

Converge Team, 2026-10-28

Most organisations do not lack data. They lack a shared picture of it. Orders live in one system, assets in another, and the spreadsheets that connect them live on someone's laptop. Every new dashboard, integration or AI assistant has to rebuild that picture from scratch, and each one gets the permissions slightly wrong.

Converge starts from the other end. It builds the picture once, as an ontology, and makes every application and every model work on it.

One model of the real world

An ontology in Converge is a set of object types (an Order, a Vehicle, a Customer), the links between them, and the actions people may take on them. Object types are backed by datasets that Data Connection keeps in sync with the systems you already run: databases, warehouses, file stores, streams and SaaS applications.

Data flows from sources, through versioned datasets, into the ontology that apps and agents share.

Once the model exists, everything else reads from it. The Objects application, maps, analytics, custom apps and the public API all see the same objects, filtered by the same permissions.

A permission set once applies everywhere the data appears, including in an AI answer.

Changes go through actions

Reading is half the story. Converge never lets an application write to a source system directly. Every change is an action: it declares its parameters, validates them, checks who is asking, and records the edit. An action can require a second person to approve it before anything happens.

Write-back to external systems follows the same rule. A governed source refuses direct pushes; writes go only through approved actions or exports, with their own grant and policy.

AI on the same terms

The AI Platform gives language models access to the ontology, not around it. Agents query objects and call functions with the identity of the person using them, and when an agent wants to change something, it proposes an action and waits. A person confirms every write.

Models themselves are chosen by role through a single gateway, so the same agent can run on a cloud model or on a model hosted inside an air-gapped site. We cover both in more detail in Deterministic by design and Cloud models, governed.

Where to start

The documentation walks through the core concepts, Data Connection, the ontology, the AI Platform and the APIs. If you already have an account, sign in at cnvrg.si.

Converge