Model Context Protocol (MCP)

Definition

Model Context Protocol (MCP) is an open standard that defines how AI models interact with external tools, data sources, and services. MCP provides a structured interface between an AI model's reasoning capabilities and real-world actions: reading data, calling APIs, executing workflows, and writing outputs.

An MCP server exposes tools with semantic names, typed parameters, and structured return values. The AI model selects which tools to invoke based on the task at hand, passes parameters, and processes results -- all through the standardized MCP interface.

Why It Matters

The primary bottleneck in enterprise AI adoption is not model capability. It is integration. Connecting an LLM to enterprise systems -- email, CRM, ERP, databases, APIs -- through custom code produces brittle, unmaintainable integrations.

MCP standardizes this integration layer. Instead of writing custom code for every AI-to-system connection, you build MCP servers that expose system capabilities as tools. The AI model discovers available tools, understands their capabilities through structured descriptions, and invokes them through a consistent protocol.

This is analogous to what REST APIs did for web services: establish a standard that makes integration composable and scalable.

How It Works

Tool Definition -- An MCP server declares its tools with names, descriptions, parameter schemas, and return types
Tool Discovery -- The AI model receives the list of available tools and their descriptions
Tool Selection -- Based on the task, the model selects which tools to invoke
Invocation -- The model provides parameters, the MCP server executes the action, and results are returned
Composition -- The model chains multiple tool invocations to accomplish complex tasks

How BAG Uses It

We operate a production platform with 68 MCP servers organized into three layers: Data CRUD (16 servers for system access), Reasoning (12 servers for analysis), and Orchestration (16 servers for multi-step workflows). This architecture enables a single AI agent to orchestrate across email, financial systems, project management, security monitoring, and more. We bring this operational experience to every client AI engagement.

Related Terms