Architecture
This page provides a high-level overview of how SmartForge transforms a user-defined business intent—created in SmartForge Studio into actionable outcomes. It shows how SmartForge Nexes acts as a data orchestration layer, unifying and coordinating data across systems, and how users can then interact with this intelligence through natural language chat or via tools exposed through an MCP server.
Build approved functions in SmartForge Studio → execute securely via SmartForge Nexes → deliver results in SmartForge Cortex or via MCP.

Core components
Section titled “Core components”SmartForge Studio is a workbench where your team builds and maintains the approved functions that SmartForge can execute.
In simple terms, SmartForge Studio defines what SmartForge is allowed to do and ensures those capabilities are clear, consistent, reviewable, and safely deployed through your existing release process.
If you only use SmartForge Cortex to ask questions or run approved actions, you typically don’t interact with SmartForge Studio directly.
SmartForge Nexes is the secure execution layer of SmartForge.
It connects SmartForge Cortex requests to approved enterprise systems and ensures every operation is validated, governed, and executed in real time.
At a high level, SmartForge Nexes is query-centric: instead of copying or syncing data, it sends queries directly to enterprise systems and merges results into a single, unified response based only on approved functions and permissions.
It:
- orchestrates requests across enterprise systems (WMS, ERP, CRM)
- validates inputs and enforces permissions (RBAC) before execution
- executes approved logic from functions
- aggregates and transforms results into a single, trusted response
SmartForge Cortex is the ** AI-powered secure conversational interface of SmartForge**.
It allows users to interact with enterprise systems using natural language while ensuring only approved functions, actions, and data access are executed.
SmartForge Cortex:
- translates natural language into approved functions (secure API calls)
- retrieves live operational data from connected enterprise systems
- supports follow-up questions with full conversation context
- formats results as tables, summaries, charts, or dashboards
SmartForge Cortex is designed for execution, not just document search, enabling real-time operational queries and approved workflows.
End-to-end request flow
Section titled “End-to-end request flow”- User request: A user asks a question in SmartForge Cortex (e.g., “Show me details of order 123”)
- Intent extraction: SmartForge interprets the request into a function + parameters
- Validation: SmartForge Nexes validates function existence, inputs, and user permissions
- Execution: Approved logic runs against enterprise systems (MOCA, REST, SQL, etc.)
- Aggregation: Results from multiple systems are merged if needed
- Response: SmartForge Cortex returns a formatted output (table, summary, chart, or dashboard)
Security model (high level)
Section titled “Security model (high level)”- Approved functions only: Only published functions can be executed
- Role-based access (RBAC): Users only access what their role allows
- Data protection: Execution stays inside your controlled environment
- Minimal LLM exposure: Only intent-related metadata is used for interpretation
- Controlled analytics: Full business data is not exposed externally; only governed metadata or optional samples (if enabled) are used for summarization