A simpler AI-facing layer can sit on top of the same ConnectUs capabilities.
REST remains the recommended interface for deterministic system-to-system integrations such as ERP, WMS, and automated order feeds. An MCP connector is a complementary option for AI assistants and operational workflows that benefit from natural language.
What is MCP in plain English?
Model Context Protocol (MCP) is a standard way for an AI application to call trusted tools. Instead of asking a user to understand URLs, HTTP methods, IDs, and JSON payloads, an AI assistant can expose higher-level actions with clear business names.
REST and MCP solve different problems
| Scenario | Recommended | Why |
|---|---|---|
| Your ERP automatically submits hundreds of orders | REST API | Deterministic, explicit, easy to test and audit |
| Your WMS synchronizes data on a schedule | REST API | Stable system-to-system contract |
| A developer builds a customer integration | REST API | Full control over validation and error handling |
| An operations user asks “Where is PO-123?” | MCP / AI | Natural language can be translated into the correct read operation |
| An AI assistant checks availability before preparing an order | MCP / AI | The connector can hide endpoint and ID complexity |
Potential business-level tools
A useful MCP layer should expose business operations rather than simply rename every REST endpoint.
search_inventory
Find available devices or SKUs using customer-friendly input.
find_projects_and_kits
Resolve available projects and kits without requiring the user to know internal IDs first.
get_kit_availability
Summarize assembled-kit availability in a form that is easy for an AI assistant to explain.
prepare_sales_order
Validate and assemble a proposed order without immediately creating it.
create_sales_order
Create a Sales Order after the caller has supplied or approved the required business data.
get_order_status
Find an order by PO reference and return current fulfillment/tracking information.
Example experience
User: Where is purchase order PO-2026-001? AI connector: 1. Searches ConnectUs Sales Orders by customer PO number. 2. Finds the ConnectUs Sales Order ID. 3. Retrieves the detailed Sales Order. 4. Returns the current status and available tracking information.
The user does not need to know that the workflow uses multiple REST calls internally.
Recommended architecture
Customer ERP / WMS ───────────────→ ConnectUs REST API AI assistant ─→ ConnectUs MCP ────→ ConnectUs API / business services
The REST API remains the stable source of truth. The MCP connector should reuse the same authentication and customer boundaries instead of creating a separate business-data model.
Safety for write operations
- Separate read tools from write tools.
- Require explicit confirmation before high-impact order creation where the AI host supports confirmation.
- Validate SKUs, quantities, addresses, and shipping choices on the ConnectUs side.
- Return structured errors to the AI instead of asking the model to guess what went wrong.
- Log tool name, authenticated customer, request ID, result, and relevant business reference.
Use the REST API documented on this portal for production customer integrations today. Treat MCP as an optional convenience layer for AI-assisted workflows rather than a replacement for the REST contract.