hitoshura25-mcp-server-generator 0.7.1

Generate dual-mode MCP servers with best practices

MCP Server Generator

A meta-generator for creating dual-mode MCP servers with best practices

Overview

Generate complete, production-ready MCP (Model Context Protocol) servers that work in two modes:

• MCP Server Mode: For AI agents (Claude Desktop, etc.)
• CLI Mode: For developers

This tool is itself an MCP server, enabling AI agents to generate other MCP servers! It demonstrates the dual-mode architecture pattern it creates and implements progressive disclosure for context-efficient tool discovery.

Why Use This?

• ⚡ Fast: Generate a complete MCP server in under 5 minutes
• 🏗️ Complete: Includes tests, documentation, packaging, and CI/CD
• ✅ Tested: Generated servers have comprehensive test suites with high coverage
• 🎯 Best Practices: Follows validated patterns from production MCP servers with built-in guidance
• 🔧 Dual-Mode: Works as both MCP server and CLI tool
• 🧠 Smart Discovery: Progressive disclosure tools for context-efficient AI agent usage
• 📦 Ready to Publish: GitHub Actions workflows included for PyPI publishing

Features

• ✅ Dual-mode architecture (MCP + CLI)
• ✅ Progressive disclosure tools (context-efficient tool discovery for AI agents)
• ✅ Built-in guidance (best practices and implementation guides)
• ✅ Claude Code integration (generate slash commands for guided development)
• ✅ Async/await support (async handlers for I/O operations, avoids event loop errors)
• ✅ Package prefix support (avoid PyPI namespace conflicts with AUTO detection)
• ✅ Complete project scaffolding (tests, docs, packaging)
• ✅ GitHub Actions workflows (via pypi-workflow-generator)
• ✅ Comprehensive test suite (92+ tests with high coverage)
• ✅ Type hints and documentation
• ✅ Best practices enforcement
• ✅ Minimal dependencies

Installation

For MCP Server Usage (Recommended)

Using uvx (no installation required):

The easiest way to use this as an MCP server - just configure in Claude Desktop:

{
  "mcpServers": {
    "mcp-server-generator": {
      "command": "uvx",
      "args": ["hitoshura25-mcp-server-generator"]
    }
  }
}

Prerequisites: Install uv:

# macOS/Linux
curl -LsSf https://astral.sh/uv/install.sh | sh

# Windows
powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

For CLI Usage (Alternative)

Using pipx (isolated installation):

pipx install hitoshura25-mcp-server-generator

Using pip:

pip install hitoshura25-mcp-server-generator

From Source (Development)

git clone https://github.com/hitoshura25/mcp-server-generator.git
cd mcp-server-generator
pip install -e .

Quick Start

Interactive Mode (Recommended)

The easiest way to get started:

hitoshura25-mcp-server-generator-cli --interactive

This will guide you through:

1. Project naming
2. Author information
3. Tool definitions
4. Configuration options

Command-Line Mode

For automation or when you have a tool definition file:

hitoshura25-mcp-server-generator-cli \
  --project-name my-mcp-tool \
  --description "Does something useful" \
  --author "Your Name" \
  --email "[e-mail]" \
  --tools-file tools.json

MCP Server Mode (For AI Agents)

Configure mcp-server-generator as an MCP server in Claude Desktop to let Claude generate MCP servers for you:

Using uvx (recommended):

{
  "mcpServers": {
    "mcp-server-generator": {
      "command": "uvx",
      "args": ["hitoshura25-mcp-server-generator"]
    }
  }
}

Using pipx/pip installation:

{
  "mcpServers": {
    "mcp-server-generator": {
      "command": "hitoshura25-mcp-server-generator"
    }
  }
}

For detailed MCP configuration, see MCP-USAGE.md

MCP Server Tools

When used as an MCP server (in Claude Desktop or other MCP clients), mcp-server-generator provides powerful tools with progressive disclosure support - allowing AI agents to discover and use tools efficiently without loading all schemas upfront.

Discovery Tools

search_tools - Find relevant tools by query

• Search by keywords, categories, or use cases
• Three detail levels for context efficiency:
  • name: Just tool names (most efficient)
  • summary: Names + descriptions + categories
  • full: Complete information including use cases
• Example: search_tools(query="generate", detail_level="summary")

get_tool_info - Get detailed information about a specific tool

• Two detail levels: summary or full
• Progressive disclosure for context efficiency
• Example: get_tool_info(tool_name="generate_mcp_server", detail_level="full")

Generation Tools

generate_mcp_server - Generate complete MCP server projects

• Creates dual-mode architecture (MCP + CLI)
• Includes tests, documentation, and CI/CD workflows
• Production-ready scaffolding with best practices
• Supports async/await patterns

generate_claude_command - Create Claude Code command files

• Generates .claude/commands/ directory structure
• Creates slash commands for guided MCP development
• Templates for common workflows: mcp_generator, best_practices, implementation_helper, custom
• Enables guided development experience

Validation Tools

validate_project_name - Validate project names

• Check Python package compatibility
• Avoid Python keyword conflicts
• Ensure PyPI naming conventions

Guidance Tools

get_best_practices - MCP development best practices

• Progressive disclosure strategies
• Context-efficient tool design
• Control flow optimization
• Security and privacy considerations
• State management patterns
• Testing strategies

get_implementation_guide - Step-by-step implementation guide

• Project setup and initialization
• Tool implementation patterns
• Testing strategies
• Deployment and publishing
• Claude Desktop integration

Why Progressive Disclosure?

Progressive disclosure allows AI agents to:

• Discover tools without loading full schemas upfront
• Save context window space for actual work
• Scale to hundreds or thousands of tools
• Get exactly the level of detail needed

Example workflow (MCP tool invocations):

# 1. Search for relevant tools
search_tools(query="generate", detail_level="name")
# Returns: ["generate_mcp_server", "generate_claude_command"]

# 2. Get summary of specific tool
get_tool_info(tool_name="generate_mcp_server", detail_level="summary")
# Returns: name, description, category

# 3. Get full details when ready to use
get_tool_info(tool_name="generate_mcp_server", detail_level="full")
# Returns: complete information including use cases and parameters

Package Prefix

To avoid namespace conflicts on PyPI, mcp-server-generator supports prefixing package names. This is highly recommended for unique package names.

Prefix Modes

AUTO (Recommended)

• Automatically detects your GitHub username from git config
• Priority: github.user → remote URL → user.name (sanitized)
• Example: my-tool → username-my-tool

Custom Prefix

• Use your own prefix (organization name, brand, etc.)
• Example: --prefix acme → acme-my-tool

NONE

• No prefix applied (only if you have a truly unique name)
• Example: unique-server-name → unique-server-name

Usage Examples

Interactive Mode:

hitoshura25-mcp-server-generator-cli --interactive
# You'll be prompted: "Prefix (default: AUTO): "
# - Press Enter for AUTO detection
# - Type "NONE" for no prefix
# - Type "acme" for custom prefix

Command-Line:

# AUTO mode (default)
hitoshura25-mcp-server-generator-cli --project-name calculator --prefix AUTO ...

# Custom prefix
hitoshura25-mcp-server-generator-cli --project-name calculator --prefix acme ...

# No prefix
hitoshura25-mcp-server-generator-cli --project-name unique-calculator --prefix NONE ...

MCP Server Mode:

{
  "project_name": "calculator",
  "prefix": "AUTO",
  ...
}

Generated Names

With prefix username and project my-tool:

• PyPI Package: username-my-tool (install with pip install username-my-tool)
• Python Import: username_my_tool (use in code as import username_my_tool)
• CLI Command: username-my-tool (run as username-my-tool --help)
• MCP Command: mcp-username-my-tool (use in config)

For detailed MCP configuration, see MCP-USAGE.md

What Gets Generated

A complete, production-ready MCP server project:

my-mcp-tool/
├── .gitignore
├── README.md
├── MCP-USAGE.md
├── LICENSE
├── setup.py
├── pyproject.toml
├── requirements.txt
├── MANIFEST.in
├── my_mcp_tool/
│   ├── __init__.py
│   ├── server.py          # MCP server implementation
│   ├── cli.py             # CLI interface
│   ├── generator.py       # Business logic (TODO stubs)
│   └── tests/
│       ├── __init__.py
│       ├── test_server.py  # MCP protocol tests
│       └── test_generator.py  # Core logic tests
└── .github/
    └── workflows/
        └── pypi-publish.yml  # PyPI publishing workflow

Generated Features

• ✅ Working MCP server with proper JSON-RPC over stdio
• ✅ CLI interface with argparse
• ✅ Complete test suite (MCP protocol + business logic)
• ✅ GitHub Actions workflow for PyPI publishing
• ✅ Comprehensive documentation (README, MCP-USAGE)
• ✅ Proper Python packaging (setup.py, pyproject.toml)
• ✅ TODO stubs for easy implementation

Tool Definition Format

Create a tools.json file to define your MCP server's tools:

{
  "tools": [
    {
      "name": "my_function",
      "description": "Does something useful",
      "parameters": [
        {
          "name": "input_text",
          "type": "string",
          "description": "Text to process",
          "required": true
        },
        {
          "name": "max_length",
          "type": "number",
          "description": "Maximum length",
          "required": false
        }
      ]
    }
  ]
}

Supported Types

• string / str
• number / int / integer / float
• boolean / bool
• array / list
• object / dict

For complete examples, see EXAMPLES.md

Documentation

• MCP-USAGE.md - Detailed MCP server configuration guide
• ASYNC_GUIDE.md - Complete guide for using async/await in generated MCP servers
• EXAMPLES.md - Example projects and use cases
• CONTRIBUTING.md - Development and contribution guidelines

Testing

The project includes a comprehensive test suite:

# Run all tests
pytest

# Run with coverage report
pytest --cov=hitoshura25_mcp_server_generator --cov-report=term-missing

# Run specific test file
pytest hitoshura25_mcp_server_generator/tests/test_server.py -v

Test Statistics:

• 92+ comprehensive tests covering all functionality
• All async MCP protocol tests passing
• Progressive disclosure and discovery tools tests passing
• Template validation tests passing

Requirements

• Python ≥3.8
• Jinja2 ≥3.0
• hitoshura25-pypi-workflow-generator ==0.6.0

Development

See CONTRIBUTING.md for detailed development instructions.

Quick setup:

# Clone the repository
git clone https://github.com/hitoshura25/mcp-server-generator.git
cd mcp-server-generator

# Create virtual environment
python3 -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

# Install in development mode
pip install -e .

# Run tests
pytest

Architecture

mcp-server-generator follows a dual-mode architecture pattern:

┌─────────────────────────────────────┐
│     mcp-server-generator            │
├─────────────────────────────────────┤
│                                     │
│  ┌──────────┐      ┌──────────┐     │
│  │ MCP Mode │      │ CLI Mode │     │
│  └────┬─────┘      └────┬─────┘     │
│       │                 │           │
│       └────────┬────────┘           │
│                │                    │
│         ┌──────▼───────┐            │
│         │ generator.py │            │
│         │ (Core Logic) │            │
│         └──────────────┘            │
│                                     │
└─────────────────────────────────────┘

Both modes use the same core generator logic, ensuring consistency.

License

Apache-2.0

Author

Vinayak Menon

Links

• PyPI: https://pypi.org/project/hitoshura25-mcp-server-generator/
• GitHub: https://github.com/hitoshura25/mcp-server-generator
• Issues: https://github.com/hitoshura25/mcp-server-generator/issues
• Reference Implementation: pypi-workflow-generator

Acknowledgments

This project is based on patterns validated in pypi-workflow-generator, a production MCP server for generating GitHub Actions workflows.

Progressive disclosure implementation follows best practices from:

• Anthropic's MCP Engineering Blog: Code Execution with MCP