Create a Connected System

Make an enterprise system AI-ready: connect it to AI Fabrix with business metadata, governed boundaries, and certification — so Role Assistants can use it safely.

This page is the end-to-end integrator walkthrough. Each step links to a phase article for depth.

Purpose

You are not only “building a connector.” You model what the business means, who may use it, what work AI may request, and how readiness is proven before AI-assisted execution.

Prerequisites

Builder CLI (aifabrix / af) on your PATH
Logged in: aifabrix auth status (run aifabrix login if needed)
Dataplane reachable for your environment
OpenAPI file/URL or MCP source for the target system

Phase flow

What this shows: The integrator path from scaffold through modeling, governance, validation, publish, certification, and operation.

What this is not: CLI command reference — use the phase articles linked below for step-by-step configure detail.

Mermaid diagram

How to set

Primary path: Quickstart CLI-first checklist. Summary:

Scaffold — aifabrix create <systemKey> --type external
System + datasource JSON — copy fixtures; set root identity (key, displayName, systemKey, entityType, resourceType, externalId)
Auth — aifabrix repair <systemKey> --auth apikey when needed
Model — field mappings, dimensions, foreign keys, exposed capabilities
Validate loop — validate → repair --dry-run → upload --probe → test ladder
Certify — verify-operations, verify-trust, verify-governance, lifecycle

Optional accelerator: AI Wizard overview — then run steps 3–6.

Defaults and examples

Minimal system + datasource pairing:

{
  "key": "example-crm",
  "displayName": "Example CRM",
  "type": "openapi",
  "authentication": {
    "credential": { "method": "apikey", "apiKey": "kv://example-crm/apiKey" }
  }
}

{
  "key": "example-customers",
  "displayName": "Customers",
  "systemKey": "example-crm",
  "entityType": "recordStorage",
  "resourceType": "customer",
  "primaryKey": ["externalId"],
  "metadataSchema": {
    "type": "object",
    "properties": { "externalId": { "type": "string", "index": true } }
  }
}

Integration folder layout:

integration/<systemKey>/
  application.yaml
  <systemKey>-system.json
  <datasourceKey>.json
  env.template
  rbac.yaml

Steps (detailed)

1. Choose the system

Pick the enterprise system that holds business reality you need — CRM, documents, tickets, HR, or custom APIs. One system key per integration folder under integration/<systemKey>/.

2. Scaffold and author JSON (CLI-first)

aifabrix create <systemKey> --type external
aifabrix resource-type list

Copy fixtures and edit system + datasource files. Optional accelerator:

aifabrix wizard
# or: aifabrix wizard --config path/to/wizard.yaml

See AI Wizard overview — wizard output still requires repair and validate.

3. Configure authentication

Many OpenAPI specs default to OAuth. For API-key systems:

aifabrix repair <systemKey> --auth apikey

Resolve secrets via kv:// paths in env.template. See Configure authentication.

4. Model business metadata

In each Business Entity, map vendor fields to business attributes — identifiers, labels, searchable fields, resource types (customer, contract, document). This is how AI gets business understanding, not raw payloads. See Configure business vocabulary.

5. Define dimensions and protection

Add dimensions (region, ownership, account) so ABAC can scope visibility. See Configure business policies.

6. Link cross-system relationships

Define foreign keys and protection rules so entities connect across Business Entities safely — contact to company, deal to customer, document to project. See Configure business policies.

7. Expose governed capabilities

Define business capabilities AI may request (customer.search, deal.reviewPipeline) — not raw API access. Wire CIP execution where transforms or sync steps are needed. See Capabilities and CIP execution.

8. Validate and repair

aifabrix validate <systemKey>
aifabrix repair <systemKey> --dry-run

Fix schema, mapping, and manifest drift before publish. See Validation and repair.

9. Test

aifabrix test <systemKey>
aifabrix test-integration <systemKey>
aifabrix test-e2e <systemKey>

Per datasource: aifabrix datasource test-e2e <datasourceKey> --app <systemKey>

10. Publish

aifabrix upload <systemKey> --probe

Publishes OpenAPI/MCP, credentials, and governance artifacts to the dataplane. See Deployment and upload.

11. Publish and certify

Enterprise AI Certification has three pillars — operations, agent metadata trust, and governance:

aifabrix verify-operations <systemKey>
aifabrix verify-trust <systemKey>
aifabrix verify-governance <systemKey> --subject-email <email>
aifabrix lifecycle <systemKey>

See Certification.

12. Operate and improve

Download, repair, and re-certify as the external system changes:

aifabrix show <systemKey> --online
aifabrix download <systemKey>
aifabrix repair <systemKey>

See Lifecycle.

Integration folder layout

integration/<systemKey>/
  application.yaml
  <systemKey>-system.json
  <systemKey>-datasource-*.json
  env.template
  <systemKey>-deploy.json

Limits

This walkthrough spans all build phases — use Minimum viable checklist when you need a single recordStorage path first. Certification pillars require live tenant subjects and protection manifests — do not skip govern steps before verify-governance.

Documentation Index