User Guides
MCP
Overview

Item MCP Overview


The Model Context Protocol (MCP) is an open standard that connects large language models (LLMs) with external tools and data sources. It allows AI systems to understand and interact with real-world context by breaking down integration barriers and eliminating data silos.

mcp-concept

What is MCP

MCP follows a client-server architecture, where a host application connects to one or more servers to access specific capabilities.

mcp-concept

General Architecture

ComponentDescription
MCP HostsApplications like Claude Desktop, IDEs, or AI chat tools that interact with MCP servers.
MCP ClientsLightweight clients that maintain direct (1:1) communication with a specific MCP server.
MCP ServersSmall, pluggable programs that expose structured capabilities (e.g., data access, task execution) via the MCP protocol.
Local Data SourcesFiles, databases, or services on your computer that MCP servers can securely access.
Remote ServicesOnline APIs or systems (e.g., e-commerce platforms, CRM, ERP) that MCP servers can connect to.

For more information, please refer to the MCP Official Documentation (opens in a new tab).

What is Item MCP

Item MCP is an open-source platform that provides a centralized hub for deploying and managing MCP servers specifically tailored for Item’s ecosystem (Item DI, OMS, WMS, BNP, etc.).

What It Offers

  • Standardized server deployment for various system connectors.
  • MCP server marketplace for quick installation of prebuilt integrations.
  • Multi-agent collaboration via AI Chat.
  • Configurable AI agents and task scheduling.
  • Secure access control for sensitive data.

Why Integrate with Item MCP

BenefitDescriptionExample
Context-Aware AIProvide real-time system context to AI modelsAutomatically link order info with shipment data
Standardized ProtocolUse a unified API interface across systemsAll agents follow the same structure
Cross-System OperationsSupport multi-agent workflowsCreate Shopify order → auto-fetch shipment info
Smart Task SchedulingAssign tasks to different agentsOne handles orders, another manages inventory
Reduced R&D CostsAdd new capabilities without touching core codeIntegrate customer support system via a new agent
Secure Access ControlOnly authorized agents access specific dataSensitive data protected by role-based permissions

Key Features

item MCP Server VS item MCP Client

AspectMCP ServerMCP Client
What it isA centralized data platform that unifies and standardizes information from multiple sources and provides secure interfaces for external systems to access data.An application that uses the MCP protocol to call data interfaces, enabling intelligent task automation and improving human-computer interaction.
Why it mattersSolves issues like data silos, delays in data flow, and cross-system collaboration challenges in industries such as warehousing and logistics.Helps users make informed decisions, facilitates customer service, and handles tasks like replenishment and exception management.
Target UsersSystem administrators and implementation teams responsible for configuring and deploying the server.Data entry teams, customer service representatives, and end-users who interact with the system.
Use Cases1) Updates and pushes OMS inventory to third-party platforms (e.g., OMS → Shopify).
2) Pushes order delivery information to third-party platforms (e.g., OMS → Shopify).
1) User requests: Querying OMS order status.
2) Automatic task triggering: Low inventory warnings, system notifications, and actionable suggestions.
Business ScenariosIntegrates with OMS and other systems to manage order fulfillment and receive updates on warehouse and transportation system statuses.Provides customer service Q&A, order status queries, and system status monitoring via mobile or web interfaces.
How it worksConfigures data access sources and permission rules, exposes standard APIs that clients call for data.Users interact through web or mobile interfaces, sending requests, receiving responses, or triggering actions.
ExampleUpdating OMS data to Shopify, managing integrations between systems.Querying order status, receiving low-inventory alerts, or handling customer service tasks.

Terminology

This glossary defines key terms you'll encounter when working with MCP clients and servers. It includes foundational concepts, communication structures, and configuration terms commonly used during setup, testing, and deployment.

Core Concepts

MCP (Model Context Protocol)
A lightweight protocol that allows AI clients to invoke structured tools provided by external servers.

MCP Server
A standalone program (Node.js or Python) that exposes callable tools using the MCP schema and responds over a supported connection type.

MCP Client An interface (e.g., AI chat, command line, dashboard) that connects to an MCP server, discovers its tools, and triggers tool executions.

Tool A unit of functionality defined by an MCP server. Each tool includes a name, description, input schema, and an execute() method.

Tool Call A request from the client to invoke a specific tool with parameters, following MCP JSON format.

Capabilities File (capabilities.json) A JSON manifest listing all tools a server provides, their schemas, and metadata. This is how clients discover what a server can do.

Prompt Template A predefined instruction that guides the AI’s response style or behavior during tool usage. Configured on the client side.

Resource A data object (e.g., file, folder, image) referenced in context but not directly executed like a tool. Resources often support agent reasoning.

Connection Types

STDIO (Standard I/O) A local connection method where the server is launched as a subprocess and communicates via standard input/output streams.

SSE (Server-Sent Events) A remote connection method where the client receives streamed JSON responses from a web-based API server.

Streamable HTTP
A type of server connection that supports continuous data exchange without closing the HTTP connection after each response, providing a more efficient way for AI model servers or integration endpoints to send updates or partial results progressively.

Configuration Parameters

Command The system command used to launch the server (required in STDIO mode), "python" or "node".

Arguments (Args) A list of arguments passed to the command (usually the server script path), ["index.js"].

Environment Variables (Env) Optional variables used to configure runtime behavior (e.g., auth paths, system tools), { "UV_PATH": "/usr/bin/uv" }.

Server URL The entry point for an SSE-based server (usually a web API endpoint), "https://api.example.com/mcp".

Headers Optional metadata or authentication passed in HTTP requests for SSE servers, { "Authorization": "Bearer token" }.

Server Type Declares the connection method (STDIO or SSE) when adding or registering a server, "stdio" or "sse".

Debugging & Testing

MCP Playground / Inspector A visual testbed for connecting to MCP servers, running tools, and viewing results in real time. Useful for debugging before deploying.

Tool History A log of previous tool calls and their outcomes (input → output). Helpful for reviewing session behavior.

Session A persistent communication state between a client and a server, allowing tools to be called with context.