Platform
Everything a business system needs. Supplied by the runtime.
Your agent describes the business in compact metadata. ObjectOS turns it into a working system: database and APIs, permission-aware screens, automated processes, approval queues, and dashboards — with governance enforced at runtime, not promised in a slide.
- 25+
- Runtime services behind every deployed application
- 6
- Capability areas, from data modeling to analytics
- ~1%
- Metadata surface the AI writes; the runtime supplies the rest
The 1% your agent writes
One definition. A whole application behind it.
This is the shape of a governed order object. Permission sets, approval flows, views, and AI tools are sibling definitions in the same reviewable stack — while tables, APIs, screens, queues, and audit come from the runtime.
import { ObjectSchema, Field } from '@objectstack/spec/data';
export const Order = ObjectSchema.create({
name: 'sales_order',
label: 'Order',
fields: {
customer: Field.lookup('crm_account', { label: 'Customer', required: true }),
total: Field.currency({ label: 'Total', min: 0 }),
discount: Field.percent({ label: 'Discount', max: 30 }),
status: Field.select({
label: 'Status',
trackHistory: true,
options: [
{ label: 'Draft', value: 'draft', default: true },
{ label: 'Submitted', value: 'submitted' },
{ label: 'Approved', value: 'approved' },
{ label: 'Fulfilled', value: 'fulfilled' },
],
}),
},
});Capability map
Six capability areas, one governed runtime
The same capabilities you would expect from a mature business platform — except your AI writes the definition and the runtime does the assembly.
Data modeling
Objects, relationships, and validations become tables, migrations, a query language, and REST APIs — on Postgres, MySQL, SQLite, MongoDB, and more.
App interfaces
Forms, views, and dashboards render straight from metadata, and every user sees exactly what their permissions allow — no hand-built frontend.
Process automation
Flows with durable pause and resume, record-change, scheduled, and API triggers, background jobs, and reliable webhook delivery.
Approvals
Multi-step approvals with user, role, team, and hierarchy resolution, escalation, and record locking while a decision is pending.
Permissions & security
Role-based, row-level, and field-level control with record sharing, tenant isolation, and an immutable audit trail.
Analytics & reporting
Aggregations, time series, funnels, and dashboards over the same governed objects — no separate BI stack to wire up.
Built for AI
Designed to be written by agents, reviewed by people
Every capability above is metadata-driven, which is exactly what makes the platform AI-writable: small definitions, readable diffs, and runtime enforcement.
AI Build & Ask
Describe a change in natural language and review the diff, or ask questions over live business data — always inside the signed-in user’s permissions.
Tools & MCP
Objects, queries, and actions become policy-checked tools for Claude, Cursor, or any MCP client — declared as metadata, never handwritten glue.
Designers & console
16+ open-source admin surfaces — the object designer, flow canvas, approvals inbox, permission matrix, and audit viewer — so people fine-tune everything the AI drafts.
The review loop
Structural changes land as compact diffs in an approval queue, so a person signs off before the runtime ships anything.
Decision surface
What changes, who reviews it, what runs
| Business need | AI writes | Runtime supplies |
|---|---|---|
| A customer database with an API | Objects, fields, relationships | Tables, migrations, query language, REST endpoints |
| Screens for each team | Views, forms, dashboards | Rendered UI that respects each user’s permissions |
| An approval before discounts ship | A flow rule with an approval step | Approval queues, escalation, record locking, audit |
| AI that answers from live data | Tool and action contracts | Policy-checked MCP tools scoped to the signed-in user |
FAQ
Questions this page should answer
Which capabilities are in the open-source edition?
The runtime capabilities on this page — data, UI, automation, approvals, permissions, analytics — are part of the open-source platform. You bring your own AI: a coding agent writes metadata as source files and any MCP client can query your objects. The in-app AI Build and Ask assistants run on Cloud and Enterprise.
Can it run in our own infrastructure?
Yes. ObjectOS is self-hostable and runs in VPCs, on local servers, or in air-gapped networks, with local models, internal identity, and your own secret management on Enterprise.
Next pages