FactVerse AI Agent Architecture
FactVerse AI Agent sits between enterprise AI clients and the FactVerse industrial runtime. It gives agents a governed way to use digital twins, operational data, knowledge, simulation services, and approved action paths while keeping product API access scoped and auditable.
Architecture layers
| Layer | Role | Main FactVerse connection |
|---|---|---|
| AI clients and channels | Enterprise copilots, agent runners, workflow tools, and customer applications that need industrial context | MCP endpoints and scoped API keys |
| Governance and tool access | Tenant boundary, scope checks, audit records, human approval, and write-action control | FactVerse Platform identity, permission, and tenant context |
| Data and knowledge | Operational signals, asset records, documents, SOPs, maintenance history, and quality checks | DFS, knowledge stores, and generated references |
| Digital twin and simulation | Scene context, equipment layout, SimReady assets, and planning models for analysis or Physical AI workflows | FactVerse Designer and simulation services |
| Operational execution | Runtime visualization, alerts, work orders, inspection records, and field feedback | DataMesh Inspector and customer operation systems |
Runtime flow
- The client connects through an MCP endpoint with a scoped credential.
- FactVerse resolves tenant, asset, permission, and product context before tool execution.
- The agent retrieves data, digital twin context, knowledge, or simulation-ready assets through governed tools.
- The agent produces analysis, recommendations, task drafts, or simulation requests based on the available context.
- Write actions such as work-order updates or operational changes stay behind approval and audit controls.
- Inspector and connected operation systems record the execution result so future agent work can learn from completed activity.
Product responsibilities
| Product or service | Responsibility in Agent workflows |
|---|---|
| FactVerse Platform | Tenant, identity, asset, permission, subscription, and shared product context |
| DFS | Data ingestion, transformation, quality checks, operational signals, and integration pipelines |
| FactVerse Designer | Digital twin scene creation, layout planning, SimReady asset preparation, and simulation-oriented authoring |
| DataMesh Inspector | Runtime visualization, alerts, inspection workflows, work orders, and field execution records |
| MCP | Governed tool discovery and access for external AI clients and enterprise agents |
| Knowledge and references | Manuals, SOPs, asset records, tool catalogs, and domain documents available to approved workflows |
Deployment models
FactVerse AI Agent can support cloud, on-premises, and hybrid deployment discussions depending on customer security and integration requirements. The architecture model separates integration boundaries, governance surfaces, and product responsibilities so deployment planning can stay clear across environments.
Design principles
- Expose capabilities through scoped tools instead of unrestricted product APIs.
- Keep tenant, asset, and permission context attached to every workflow.
- Treat digital twins as operational context with visual, data, and workflow meaning.
- Use simulation and planning assets where they improve reasoning quality or scenario validation.
- Keep human approval and audit records around actions that affect real operations.