Vavan AI
Vavan Core

Vavan Core

Vavan Core is Vavan's shared, typed semantic model of a business — one model that every app and AI agent reads and writes.

Vavan Core is a typed model of the real things a business operates on and the relationships between them. Vavan Core is the center of the platform: a single, shared representation of your customers, sites, orders, drivers, products, and the connections among them. Every app and every AI agent reads and writes the same objects through it.

The alternative — the default for most B2B operators — is the opposite of a shared model. A CRM has its own idea of an "account." The ERP has a separate "customer." A routing tool has yet another "stop" record. Each system carries a partial, slightly different copy of the same reality, and keeping them in sync means a thicket of point-to-point integrations that drift apart over time. Vavan Core replaces that with one model: when something is true about a customer, it is true everywhere, because there is only one customer object.

Note. Vavan Core is defined as configuration, not hardcoded into each app. Apps don't own their own schemas — they share Core. That is what removes per-app schema drift and lets data and AI flow across the platform without bespoke integrations.

The four building blocks

Vavan Core is built from four kinds of definitions. Together they describe what exists, what is known about it, how it connects, and what can be done to it.

Object types

An object type is a kind of real-world thing — an Account, a Contact, a Product, an Order, a Site, a Driver, a Vehicle, a Route, a Shipment, a Signal, a Task. Each object type has a name, a set of properties, and links to other types. Concrete records (a specific customer, a specific order) are instances of these types.

Properties

Properties are the typed fields on an object type. Each property has a type — text, number, money, date, enum, boolean, geo — and some are derived or computed from other data rather than stored directly. Typing the fields is what makes the model safe for apps and AI to reason over.

Link types

Link types are the typed relationships between object types: an Account has many Contacts, an Order belongs to one Account, an Order has Line Items that reference Products, a Route is made of Stops that point at Orders. Links turn a pile of records into a navigable graph of your business.

Action types

Action types are the governed writes that change Vavan Core — creating a contact, advancing a pipeline stage, dispatching an order, enrolling a contact in a sequence. Instead of arbitrary updates scattered across apps, mutations go through defined, permissioned actions, so every change is consistent and auditable.

A config-driven model

Object types, properties, and links are defined as configuration, not baked into application code. Adding a property or a new object type is a change to the model, not a migration that each app has to learn about separately. Because every app reads and writes against the same configured definitions, there is no per-app schema drift and no point-to-point integration to maintain between apps.

In production, an organization's Vavan Core typically spans on the order of dozens of object types and hundreds of properties. The exact shape grows with the business and the systems connected to it — the model is meant to be extended, not frozen.

How data is normalized into Vavan Core

Vavan Core is not a place you re-enter data by hand. Records are normalized into it from the systems you already run — Salesforce, HubSpot, your ERP, accounting, email. Incoming records are mapped onto the right object types and properties, resolved against existing objects so one real company stays one object, and kept current as the source systems change.

See Platform architecture for how the connect-and-normalize pipeline works, and Integrations for the systems you can connect.

How apps and AI consume it

Every Vavan application — CRM, Dispatch, Maps, Stores — reads and writes the same Vavan Core objects. An Account edited in the CRM is the same Account a driver sees on a Dispatch route. A signal raised against an Order can become a Task linked to the same Account, surfaced in a different app, with no integration in between.

Because the model is typed and explicit, AI agents can reason over it reliably: they know what an Order is, what it links to, and which action types are allowed to change it. AI mutates Vavan Core through the same governed action types as the apps, so automated changes carry the same constraints and auditability as human ones.

Organizations and isolation

Vavan Core is multi-organization. Every object — and every property, link, and action on it — is scoped to an organization. Row-level security enforces that one organization can never read or write another organization's objects, even though they share the same platform and the same object-type definitions. Your model and your data stay yours.

See Security & isolation for how isolation and row-level security are enforced.

Building blocks

Continue into each building block for the full reference:

Building block What it is Reference
Object types The kinds of real-world things you operate on — Account, Contact, Product, Order, Site, Driver, Route, Signal, Task. /docs/ontology/object-types
Properties The typed fields on an object type — text, number, money, date, enum, boolean, geo, and derived/computed values. /docs/ontology/properties
Link types The typed relationships between object types — Account to its Contacts, Order to an Account, Route through its Stops. /docs/ontology/link-types
Action types Governed writes that mutate Vavan Core — create a contact, advance a pipeline stage, dispatch an order. /docs/ontology/action-types