ZaVang/prompt-formatter-mcp

Prompt Formatter MCP Server

MCP server for intelligently formatting and optimizing prompts using rule-based analysis and structured templates to enhance LLM performance across different models.

Core Philosophy

• Convention over Configuration: Follow our for best results
• Rule-Based Processing: Fast, deterministic formatting without external LLM API calls
• Agent Collaboration: MCP handles structure, Agent handles intelligent optimization
• Template-Driven: Use default templates or bring your own

Features

✅ Phase 1 (MVP - In Development)

• 🎯 Smart Content Analysis: Rule-based detection of prompt sections (introduction, rules, examples, etc.)
• 📝 Default Templates:
  • Claude XML format (with semantic tags like <introduction>, <rules>, <section>)
  • GPT Markdown format (with headers and structured lists)
  • JSON format for structured data
• 🔍 Quality Analysis: Evaluate prompt completeness and clarity
• ✅ Convention Validation: Check if prompts follow best practices
• 🎨 Custom Templates: Define your own template structure

🚧 Phase 2 (Planned)

• 📚 Template library with pre-built prompt patterns
• 🔄 Batch processing support
• 📊 Performance analytics

💭 Phase 3 (Future)

• 🧠 Learning from feedback
• 🌐 Web interface
• 🔌 VSCode extension

MCP Tools

1. format_prompt

Format raw prompts into structured templates:

• Input: Raw prompt text
• Output: Formatted prompt in XML/Markdown/JSON/Custom format
• Features: Section detection, template application, confidence scoring

2. analyze_prompt

Analyze prompt quality and structure:

• Input: Raw prompt text
• Output: Detected sections, quality metrics, improvement suggestions
• Purpose: Returns analysis for Agent to use in optimization

3. validate_prompt

Validate prompt against writing conventions:

• Input: Raw prompt text
• Output: Compliance score, validation checks, recommendations
• Purpose: Help users write better prompts

Quick Start

Installation

# Clone the repository
git clone https://github.com/ZaVang/prompt-formatter-mcp.git
cd prompt-formatter-mcp

# Install dependencies
npm install

# Build the project
npm run build

Setup with Claude Desktop

1. Edit your Claude Desktop configuration file:

   • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
   • Windows: %APPDATA%\Claude\claude_desktop_config.json
   • Linux: ~/.config/Claude/claude_desktop_config.json

2. Add the MCP server configuration:

{
  "mcpServers": {
    "prompt-formatter": {
      "command": "node",
      "args": ["/absolute/path/to/prompt-formatter-mcp/dist/index.js"]
    }
  }
}

3. Restart Claude Desktop

Basic Usage

Once configured, you can use the MCP tools in Claude Desktop:

Format a prompt:

Please use the format_prompt tool to format this prompt:
"You are a translator. Rules: 1. Keep formatting 2. Use natural language. Task: Translate this text."

Analyze a prompt:

Please use the analyze_prompt tool to analyze this prompt and suggest improvements:
"Translate this to Chinese"

Validate a prompt:

Please use the validate_prompt tool to check if my prompt follows best practices:
"You are a translator..."

For detailed usage instructions, see

Documentation

Getting Started

• 📖 - Complete guide on installation, setup, and using all MCP tools
• ✍️ - Start here! Best practices for writing prompts
• 🚀 - Get up and running in minutes

Architecture & Design

• 🏗️ - Detailed architecture and implementation plan
• 🔧 - Understanding MCP server structure and patterns

Advanced Topics

• 🎨 - Create your own formatting templates
• 🧪 - How to test the MCP server
• 📊 - Tips for optimizing prompts for different models

Development

• ⚡ - Recent improvements and features
• ✅ - MVP milestone achievements

Why Prompt Formatter?

Different LLMs have different format preferences:

• Claude performs better with XML-structured prompts with semantic tags
• GPT models prefer Markdown with clear sections
• All models benefit from well-organized prompts with clear examples

This MCP server helps you:

1. Write prompts following best practices (see CONVENTIONS.md)
2. Automatically format them for your target model
3. Validate and improve prompt quality
4. Let the Agent optimize based on structured analysis

License

MIT License - see LICENSE file for details.