Platform · Architecture
One architecture, every layer.
Business applications, AI platform agents and the data platform all run on the same platform services, over the same infrastructure.
Every layer above runs the same way for every tenant — isolation happens at the infrastructure layer, not by forking code per tenant.
The five layers.
ERP Platform
The single platform every tenant runs on — one set of engines, interpreting metadata per tenant.
Business Apps · AI Platform Agents · Data Platform
What runs on the platform: the business applications, AI agents that act on the same data, and the analytics/data layer reading it live.
Platform Services
The shared services every application and agent calls into rather than reimplementing.
Identity · Workflow · Integration
The three platform services underneath everything: who you are, what process you are in, and what systems you are connected to.
Infrastructure
Stateless compute, per-tenant databases and object storage, scaled independently of the application layer.
What each layer does.
Business Apps
Finance, CRM, HCM, Supply Chain, Manufacturing, Service and every other Spark ERP application — each one built from the same page, form, workflow and entity metadata, interpreted by the same engines.
AI Platform Agents
AI agents sit next to the business applications, not above them — they read and act through the same entity model, workflow engine and permission checks a human user would go through.
Data Platform
Analytics, dashboards and reporting read the same live entity model the applications write to, so there is no separate reporting copy of the data to reconcile.
Platform Services
Identity (authentication, SSO, MFA, RBAC), Workflow (stateful approvals and processes) and Integration (the connector framework and API orchestration) are implemented once and shared by every application and agent above them.
Infrastructure
The application layer is stateless — any instance can serve any request — so it scales horizontally behind a load balancer. Each tenant's database and schemas sit underneath, isolated at the connection layer.
See how tenants are isolated at every layer.
The Multi-Tenant Architecture page goes deeper on database and schema isolation, regional deployment, and disaster recovery.