Getting started
Overview
Get oriented to the Fabric platform — what it is, how the docs are organized, and where to go next.
Fabric is an ontology-based, agent-native runtime for line-of-business applications. It owns the mechanics of governed mutation: actions, policies, state machines, events, projections. It owns no industry vocabulary.
You build a Fabric app by:
- Declaring your domain ontology — object types, actions, policies, events, state machines — in a
FabricModule. - Composing one or more modules into a host app.
- Letting every domain mutation flow through one pipeline.
This site is the reference manual for that runtime.
How the docs are organized
| Section | What it answers |
|---|---|
| Mental model | One-page glossary — what each ontology concept means. |
| Quickstart | Install, declare a module, fire an action, observe the event. |
| Platform reference | The conceptual / reference manual: architecture, action pipeline, entities, governance, patterns. |
The one diagram to start with
Every change to domain state in a Fabric app passes through this. There are no exceptions, only a fixed list of actor types (natural_person, agent, system, service_account, external_system, integration) and three valid call paths (triggers).
What's new about it
If you've worked with event-sourced systems, most of this will feel familiar. The novelties:
- Agents are first-class actors. The same pipeline that handles your operational team handles your AI agents. No side door, no separate audit log.
- Policies are versioned and auditable. Each policy decision carries
dispatchEvidencerecording exactly why it returnedpass/warn/block. - The platform is portable; verticals aren't.
@fabric/platformhas no DB client, no ORM, no industry vocabulary. Verticals bind concrete clients and own the ontology.
Next
- New to event-sourced platforms? Read the mental model first.
- Want to see code? Jump to quickstart.
- Want the architecture? Open platform reference.