Stage 5: BUILD

Scaffold and implement MCP server/Apify Actor from specification.

Prerequisites

Requires outputs/spec/{name}-spec.md from Stage 4.

Parameters

Parse from user input:

• --name: Spec name to build (required if multiple specs exist)
• --target: apify | mcp-standalone | npm (default: apify)

Process

Use the mcp-builder agent to:

1. Load spec from outputs/spec/{name}-spec.md

2. Scaffold Apify Actor structure:

   outputs/build/{name}/
   ├── src/
   │   ├── main.ts          # Entry point
   │   ├── routes.ts        # Request handlers (if applicable)
   │   └── utils/
   │       ├── api.ts       # External API clients
   │       └── helpers.ts   # Utility functions
   ├── .actor/
   │   ├── actor.json       # Actor configuration
   │   └── input_schema.json # Input validation schema
   ├── package.json
   ├── tsconfig.json
   └── README.md
   

3. Implement core functionality:

   a. Main entry point with Apify Actor lifecycle

   import { Actor } from 'apify';
   
   await Actor.init();
   try {
     const input = await Actor.getInput();
     // Implementation
     await Actor.charge('event-type', { count: n });
   } finally {
     await Actor.exit();
   }
   

   b. PPE charging for each billable event

   c. Input schema with validation

   d. Error handling with proper logging

   e. Progress tracking for long operations

4. Generate input_schema.json from spec requirements

5. Generate README.md with usage examples

6. Save complete codebase to outputs/build/{name}/

Output Structure

outputs/build/{name}/
├── src/
│   ├── main.ts
│   ├── routes.ts
│   └── utils/
├── .actor/
│   ├── actor.json
│   └── input_schema.json
├── package.json
├── tsconfig.json
├── README.md
└── .gitignore

Actor.json Template

{
  "actorSpecification": 1,
  "name": "{name}",
  "title": "{title}",
  "description": "{description}",
  "version": "0.1.0",
  "buildTag": "latest",
  "input": "./input_schema.json",
  "dockerfile": "./Dockerfile",
  "storages": {
    "dataset": {
      "actorSpecification": 1,
      "title": "Output dataset",
      "description": "Results from the actor run"
    }
  }
}

Next Step

Run /mcp-pipeline:qa to test the implementation.