Skip to main content

​
MCP Server

​
Overview

ORBEXA includes a full MCP Server implementation (the OrbexaMCPServer class, defined in mcpServer.ts), providing AI agents with a standardized interface for tool invocations and resource access. AI agents can query products, inventory, and categories directly through MCP tools without needing to understand the underlying database structure.

​
5 MCP Tools

​
search_products

Product search tool supporting multi-dimensional search by keyword, category, price range, and more. Input parameters:
  • query — Search keywords
  • category — Product category (optional)
  • min_price / max_price — Price range (optional)
  • limit — Maximum number of results (optional)
Output: A list of matching products including name, price, inventory status, and product ID.

​
get_product

Retrieve full details for a single product. Input parameters:
  • product_id — Product ID
Output: Complete product information including description, variants, images, and specifications.

​
check_availability

Query product inventory availability. Input parameters:
  • product_id — Product ID
  • variant_id — Variant ID (optional)
Output: Inventory quantity and availability status.

​
get_categories

Retrieve the merchant’s product category listing. Input parameters: No required parameters. Output: Category tree structure with category names and product counts.

​
get_brands

Retrieve the merchant’s brand listing. Input parameters: No required parameters. Output: Brand list with associated product counts.

​
3 MCP Resources

MCP resources provide read-only data access. AI agents can subscribe to resources for real-time updates.

​
ucp://inventory/:merchant_id

Real-time inventory data for a merchant. AI agents subscribing to this resource receive inventory change notifications.

​
ucp://catalog/:merchant_id

The merchant’s complete product catalog. Contains basic information for all published products.

​
ucp://schema/ucp-standard

The UCP standard Schema definition. AI agents can use this resource to understand the UCP data format specification.

​
HTTP API Routes

The MCP Server is exposed externally through the following HTTP endpoints:
EndpointFunction
/api/mcp/toolsLists all available tools with their input/output schemas
/api/mcp/executeExecutes a tool call with the tool name and parameters
/api/mcp/resourcesLists all available resources and their URIs
/api/mcp/usageQueries API usage statistics
/api/mcp/promptsRetrieves predefined prompt templates

​
Authentication

All MCP endpoints require API key authentication. AI agents include the API key in the request header: 
Authorization: Bearer YOUR_API_KEY
API keys are automatically generated when a merchant connects (Shopify/WooCommerce/WordPress), or can be created manually in the console.

​
Custom Domain Support

ORBEXA supports accessing MCP endpoints through the merchant’s custom domain. Implemented via mcpDomainRouter, merchants can bind MCP services to their own domain: 
https://shop.example.com/api/mcp/tools
Rather than being limited to the ORBEXA platform domain. This allows AI agents to call MCP tools directly after discovering a merchant’s domain.

​
NPM Package: OTR MCP Tools

In addition to the merchant-level MCP Server, ORBEXA publishes an independent NPM package @otr-protocol/mcp-server that provides MCP tools for the OTR trust registry:

​
verify_merchant

Verify a merchant’s trust score. Input: Merchant domain Output: OTR trust score, six-dimension breakdown, and trust badge

​
search_registry

Search the OTR trust registry. Input: Search keywords Output: List of matching registered merchants with their trust scores This NPM package can be installed independently in any MCP client (such as Claude Desktop) to query the OTR trust registry without requiring an ORBEXA merchant account.

​
Summary

The ORBEXA MCP Server provides 5 tools and 3 resources covering product search, detail retrieval, inventory checking, and category and brand browsing. API key authentication ensures security, custom domain support enhances flexibility, and the independent NPM package allows any AI agent to query the OTR trust registry.
Next chapter: Data Cleaning and Quality — AI Refinery Pipeline, flywheel learning, and HITL review