​

Model Context Protocol Overview

The Model Context Protocol (MCP) is a standardized system for defining, validating, and interacting with AI model contexts across the Zopio codebase. It provides a consistent interface for context management, enabling seamless communication between different parts of your application and AI models.

​

What is MCP?

MCP is designed to solve the challenge of providing structured, validated context to AI models in a consistent way. It implements a resource-based architecture that allows you to:

1. Define context resources with strict schemas
2. Validate resources against those schemas
3. Share resources between different parts of your application
4. Consume resources in AI model interactions

​

Key Concepts

​

Resources

Resources are the core building blocks of MCP. Each resource:

• Has a unique identifier
• Belongs to a specific type
• Contains attributes that follow a defined schema
• Can be validated against its schema

Example resource:

{
  "id": "user-123",
  "type": "user",
  "attributes": {
    "name": "John Doe",
    "email": "john@example.com",
    "role": "admin"
  }
}

​

Schemas

Schemas define the structure and validation rules for resources. MCP uses Zod for schema definition and validation, providing:

• Type safety
• Runtime validation
• Detailed error messages
• TypeScript integration

​

Server/Client Architecture

MCP implements a server/client architecture:

• Servers host resources and provide endpoints for clients to access them
• Clients consume resources from servers and use them in AI interactions

This architecture allows for flexible deployment options, from in-memory servers for testing to full API-based servers for production.

​

When to Use MCP

MCP is particularly useful when:

• You need to provide structured context to AI models
• You want to ensure context data is validated before use
• You need to share context between different parts of your application
• You’re building AI features that require consistent context formats

​

Getting Started

To start using MCP in your project:

# Install the package
pnpm add @repo/mcp

Then import the necessary components:

// Import core components
import { MCPServer, MCPClient, createResource } from '@repo/mcp';

// Import schemas as needed
import { modelSchema, apiSchema } from '@repo/mcp';

See the other sections in this documentation for detailed usage examples and API reference.