ObjectOSBuild
How apps come to life in ObjectOS — by chatting with AI, by clicking through forms in Studio, or by forking a template.
Building in ObjectOS is authoring your ontology — its core (objects and fields, relations, actions, permissions, flows, agents) and the projections people see it through (views, dashboards, apps). Most people don't write that metadata by hand: they describe what they want and the AI Builder writes it. Three paths exist, in order of how most people actually use the platform:
| Path | Who's it for | Output |
|---|---|---|
| AI Builder (primary) | Business users, product owners, anyone | Live metadata in your tenant, queued for approval |
| Click-build (fallback) | Admins who prefer Studio's forms | The same live metadata |
| Template fork (dev path) | Engineers who want code under source control | A TypeScript package shipped via the marketplace |
All three paths author the same ontology — the same objects, fields,
relations, actions, permissions and flows, with the same projections — and
every path passes the same validation gate before anything runs: the
@objectstack/spec schemas and the authoring rules os validate runs,
which the runtime also applies to every metadata write from the AI
Builder or Studio. See
Validating Metadata
in the ObjectStack docs.
The platform vocabulary
Every app is composed from the same primitives. Learn these once.
| Concept | What it is | Doc |
|---|---|---|
| Package | The unit of organization — com.acme.crm, versioned, installable | Packages |
| Data model | Objects + fields + relationships + state machines | Data Model |
| Interface | Apps, views, forms, dashboards, and pages — what your users see | Interface |
| Actions | Named operations — callable from REST and from buttons, and by AI agents once you expose them; one can start a flow | Actions |
| Permissions | Who may read, create, edit and delete what — permission sets, record access, field-level security | Permissions |
| Flows | Declarative business logic (autolaunched / scheduled / manual) | Flows |
| Agents | End-user AI assistants — Agent → Skill → Tool | Agents |
| Marketplace | Ready-made packages anyone can install in one click | Marketplace |
Where to start
-
Just exploring? → Open ObjectOS, start the AI Builder, say "I need to track customer support tickets with priority, status, and assignee." You'll have a working object in under 30 seconds.
-
Building a real product on top? → Read Packages and Data Model so you understand what the AI is generating. Use the AI to draft, refine by hand or by chat, commit the result.
-
Replacing a Salesforce/Retool app? → Fork the closest marketplace template, strip what you don't need, ask the AI to fill in the gaps.
What you'll never have to do
- Write a REST endpoint. The platform generates one per object.
- Wire a Studio form. The form renders from the schema.
- Write RBAC checks. Permissions are declared, not coded.
- Build an admin UI. Setup and Account cover system administration and sign-in.
- Generate types.
@objectstack/specderives them from the schema.
ObjectOS is opinionated about this on purpose: once metadata exists, every surface (REST, the UI, ObjectQL, audit, AI tools) appears for free. That's why the AI Builder works — it only needs to produce metadata, and the platform handles the rest.
Profile & settings
Make ObjectOS yours — your name and avatar, your theme and language, and control over every signed-in session.
AI Builder
The built-in chat that turns plain-language requirements into running metadata, and the assistant that answers questions about your data — how each works, what tools they call, and how changes are approved.