First MCP: Using MCP for Tech Stack Selection

In this tutorial, you'll use project_builder — one of Prime Style MCP's core prompts — to experience the tech stack selection process for a new web application.

🎯 Goals

• Understand the basics of using MCP.
• Use the project_builder prompt to receive tech stack recommendations based on project requirements.
• Learn how to interpret MCP responses and turn them into action.

Fictional Scenario

You're a member of a startup engineering team. Your team has decided to build a simple internal blog system to streamline knowledge sharing.

In this tutorial, you'll consult MCP about tech stack selection using this scenario.

Step 1: Organize Requirements

First, clarify the requirements for the application you want to build. The project_builder prompt accepts a task_description parameter where you describe the requirements.

Here's a simple set of requirements for this scenario:

- Support posting articles in Markdown
- Allow comments on articles
- Authenticate with Google accounts
- Development team: 3 people (1 frontend, 2 backend)
- Target development period: 3 months
- Infrastructure: AWS

Step 2: Run the project_builder Prompt

There are two main ways to invoke MCP prompts from Claude Code.

Method 1: Slash Command (Recommended)

When you know the prompt name, use a slash command for the most reliable invocation. Start claude and enter the following. Parameters follow the command.

/mcp__psm__project_builder task_description="Internal blog system.\n- Markdown article posting\n- Article comments\n- Google account authentication\n- Team of 3 (1 frontend, 2 backend)\n- 3-month timeline\n- AWS infrastructure"

• psm is the name you specified when registering the server.
• Pass arguments as parameter_name="value".

Method 2: Mention

You can also mention with @<server-name> and make a natural language request:

@psm Use the project_builder prompt to suggest a tech stack for an internal blog system.

Requirements:
- Markdown article posting
- Article comments
- Google account authentication
- Team of 3 (1 frontend, 2 backend)
- 3-month timeline
- AWS infrastructure

Claude understands the context and translates it into the appropriate parameters.

Reference: Calling the API directly with curl

The MCP server also provides standard HTTP endpoints. If Claude Code is unavailable or you want to call it from scripts:

# Requires MCP_SERVER_URL and MCP_API_KEY environment variables
curl -X POST "$MCP_SERVER_URL/prompts/project_builder" \
  -H "Authorization: Bearer $MCP_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "task_description": "Internal blog system.\n- Markdown article posting\n- Article comments\n- Google account authentication\n- Team of 3\n- 3-month timeline\n- AWS infrastructure"
  }'

Step 3: Interpret the Results

MCP returns a report with several tech stack options and their pros and cons based on your requirements. (Sample response below)

{
  "project_id": "proj_internal_blog_01",
  "summary": "Tech stack proposal for internal blog system",
  "options": [
    {
      "name": "React (Next.js) + Python (FastAPI)",
      "pros": [
        "Clear separation of frontend and backend — easy to divide work.",
        "SSR via Next.js for fast initial load.",
        "FastAPI's async support provides high performance."
      ],
      "cons": [
        "Requires knowledge of both frontend and backend.",
        "Management overhead may be higher for a small team."
      ],
      "recommended_for": "Best for teams prioritizing modern development practices with future scalability in mind."
    },
    {
      "name": "Python (Django) + HTMX",
      "pros": [
        "Monolithic architecture — backend engineers can lead development.",
        "Relatively low learning curve.",
        "Rapid prototyping possible."
      ],
      "cons": [
        "Not suited for complex frontend interactions.",
        "Rich SPA-like UI/UX is difficult to achieve."
      ],
      "recommended_for": "Best for backend-focused teams wanting to ship an MVP quickly."
    }
  ],
  "recommendation": "Given team composition and timeline, Python (Django) + HTMX carries the lowest risk and offers the fastest ramp-up. If a richer UI is needed in the future, consider refactoring to split the frontend."
}

Use these results as material for team discussion. MCP is an assistant that provides data for decision-making — the final call should be made by the team based on skills and product vision.

Summary

In this tutorial, you experienced using project_builder to get concrete tech stack options from project requirements.

Other MCP prompts similarly structure your tasks and support decision-making. Check the Features page to explore more capabilities.