What is MCP

MCP (Model Context Protocol) is a standardized protocol for connecting AI tools to external data sources and tools. Proposed by Anthropic, it provides AI models with a standardized way to invoke tools.

Why MCP Matters

Problems with Traditional AI Tool Calling

Tool usage instructions need to be included in every prompt
AI may misunderstand tool parameters
Lack of structured return formats
Difficult to ensure call security and reliability

MCP Advantages

Standardization: Unified interface definitions
Structure: Clear input/output formats
Verifiability: Parameter type checking
Extensibility: Easy to add new tools

AgileBuilder MCP Architecture

┌─────────────┐
│   AI Tool   │
│ Cursor/Claude│
└──────┬──────┘
       │ MCP Protocol
┌──────▼──────┐
│ MCP Client  │
└──────┬──────┘
       │ HTTP
┌──────▼──────┐
│ AgileBuilder │
│  MCP Server │
└──────┬──────┘
       │
┌──────▼──────┐
│ Local Data  │
│Templates/Docs│
└─────────────┘

AgileBuilder MCP Tools

listTemplates

List all available project templates.

// Request
{
  "tool": "listTemplates",
  "parameters": {}
}

// Response
{
  "templates": [
    {
      "name": "react-starter",
      "description": "React project starter template"
    }
  ]
}

getTemplateInfo

Get detailed template information.

// Request
{
  "tool": "getTemplateInfo",
  "parameters": {
    "template": "react-starter"
  }
}

// Response
{
  "name": "react-starter",
  "description": "React project starter template",
  "variables": [
    {
      "name": "projectName",
      "type": "string",
      "required": true
    }
  ]
}

generateProject

Generate a project.

// Request
{
  "tool": "generateProject",
  "parameters": {
    "template": "react-starter",
    "output": "./my-app",
    "variables": {
      "projectName": "my-app"
    }
  }
}

// Response
{
  "success": true,
  "projectPath": "./my-app",
  "message": "Project generated successfully"
}

MCP Resources

Beyond tool calls, MCP also supports Resources, which provide contextual information to AI.

Document Resources

resource://agile-builder/docs/{doc-slug}

AI can access through Resources:

Team coding standards
API design guidelines
Naming conventions
Best practices documentation

Template Resources

resource://agile-builder/templates/{template-name}

AI can access through Resources:

Template structure preview
Variable definitions
Generation examples

How It Works

Start the MCP service: AgileBuilder starts a local MCP server
Configure the AI tool: Set up the MCP service address in your AI tool
AI calls tools: User makes a request, AI automatically calls MCP tools
Return structured results: AI processes tool results for next steps

vs. Traditional Prompts

	Traditional Prompts	MCP
Tool Description	In prompt	Standardized interface
Parameter Passing	Natural language	Structured parameters
Return Format	Natural language	Structured data
Accuracy	Depends on AI understanding	Type-safe
Reliability	May vary	Deterministic

Next Steps

MCP Service Configuration - Configure MCP service
AI Workflow - Configure AI workflows
AI Use Cases - Practical examples