modelctx

modelctxは、Pythonでのモデル管理を効率化するためのツールです。ユーザーは、モデルのバージョン管理や依存関係の追跡を簡単に行うことができ、開発プロセスをスムーズにします。特に、機械学習プロジェクトにおいて、モデルの再現性を確保するための機能が充実しています。

GitHubスター

8

ユーザー評価

未評価

お気に入り

0

閲覧数

53

フォーク

0

イシュー

0

README
ModelContextKit

🚀 A comprehensive CLI tool for rapidly creating, configuring, and deploying Model Context Protocol (MCP) servers

Python
License
Website

🌐 Visit our website: modelcontextkit.com

What is MCP?

The Model Context Protocol (MCP) is an open standard developed by Anthropic that enables AI applications (like Claude) to connect with external tools, data sources, and APIs in a standardized way. Think of it as "USB-C for AI applications" - it provides a universal connector between AI models and external systems.

Features

Quick Setup: Generate production-ready MCP servers in minutes, not hours
🔧 Multiple Backends: Support for databases, APIs, filesystems, web scrapers, email, and cloud storage
🎯 Interactive Wizard: Guided setup with intelligent configuration prompts
📚 Comprehensive Templates: Pre-built templates following MCP best practices
🔒 Security First: Built-in security features and input validation
🧪 Testing Ready: Generated projects include test suites and examples
📖 Auto Documentation: Comprehensive documentation generation
🎨 Claude Desktop Integration: Automatic configuration for Claude Desktop
Production Ready: Generates syntactically correct, tested code that works out of the box

Recent Improvements

🔧 Fixed Code Generation: Resolved indentation and syntax issues in generated MCP servers
📋 Corrected MCP Schema: Updated to use proper input_schema format
🧪 Improved Tests: Updated test templates to use modern MCP patterns
🔗 Fixed Resource URIs: Simplified resource implementations to avoid template issues
Enhanced Reliability: All backend types now generate working, compilable code

Installation
# Install base package
pip install modelcontextkit

# Install with specific backend support
pip install modelcontextkit[database]
pip install modelcontextkit[api]
pip install modelcontextkit[filesystem]
pip install modelcontextkit[webscraper]
pip install modelcontextkit[email]
pip install modelcontextkit[cloudstorage]

# Install with all backends
pip install modelcontextkit[all]
Quick Start
Interactive Wizard (Recommended)
modelctx wizard
Command Line Interface
# List available backend types
modelctx list

# Create a database MCP server
modelctx create my-db-server --backend database

# Create an API integration server
modelctx create my-api-server --backend api

# View templates
modelctx templates
Supported Backends
🗄️ Database Backend

Connect to SQL databases (PostgreSQL, MySQL, SQLite) with built-in connection pooling and security.

Generated Tools:

  • execute_query(query: str) - Execute SQL queries safely
  • get_table_schema(table_name: str) - Get table structure
  • list_tables() - List all tables

Configuration:

  • Database connection parameters
  • Connection pooling settings
  • Query timeout and limits
🌐 REST API Backend

Integrate with REST APIs with authentication, rate limiting, and error handling.

Generated Tools:

  • api_request(endpoint: str, method: str, data: dict) - Make HTTP requests
  • get_api_status() - Check API health

Authentication Support:

  • Bearer tokens
  • API keys
  • OAuth2 flows
📁 Filesystem Backend

Access and manipulate local files and directories with security controls.

Generated Tools:

  • read_file(file_path: str) - Read file contents
  • write_file(file_path: str, content: str) - Write to files
  • list_directory(dir_path: str) - List directory contents
  • search_files(pattern: str, directory: str) - Search for files
🕷️ Web Scraper Backend

Scrape and parse web content with respect for robots.txt and rate limiting.

Generated Tools:

  • scrape_url(url: str) - Extract content from web pages
  • extract_links(url: str) - Get all links from a page
  • take_screenshot(url: str) - Capture page screenshots
📧 Email Backend

Send and receive emails via SMTP/IMAP with support for attachments.

Generated Tools:

  • send_email(to: str, subject: str, body: str) - Send emails
  • list_emails(folder: str, limit: int) - List emails from folder
  • read_email(email_id: str) - Read specific email
☁️ Cloud Storage Backend

Connect to cloud storage services (AWS S3, Google Cloud Storage, Azure Blob).

Generated Tools:

  • upload_file(local_path: str, remote_key: str) - Upload files
  • download_file(remote_key: str, local_path: str) - Download files
  • list_objects(prefix: str) - List stored objects
  • delete_object(key: str) - Delete objects
Generated Project Structure
my-mcp-server/
├── server.py                 # Main MCP server file
├── requirements.txt          # Python dependencies
├── pyproject.toml           # Python project metadata
├── README.md                # Project documentation
├── .env.template            # Environment variables template
├── .gitignore              # Git ignore rules
├── config/
│   ├── config.yaml         # Main configuration file
│   ├── claude_desktop_config.json  # Claude Desktop integration
│   └── logging.yaml        # Logging configuration
├── src/
│   ├── __init__.py
│   ├── models/             # Data models
│   ├── services/           # Business logic
│   └── utils/              # Utility functions
├── tests/
│   ├── __init__.py
│   ├── test_server.py      # Server tests
│   └── test_tools.py       # Tool-specific tests
├── docs/
│   ├── API.md              # API documentation
│   └── DEPLOYMENT.md       # Deployment guide
└── scripts/
    ├── setup.sh            # Setup script
    └── deploy.sh           # Deployment script
Configuration

The tool supports flexible configuration through:

  • Interactive prompts during wizard mode
  • Configuration files (YAML/JSON)
  • Environment variables for sensitive data
  • Command-line arguments for automation

Example configuration:

# config/config.yaml
server:
  name: "my-api-server"
  description: "API integration MCP server"
  
backend:
  type: "api"
  base_url: "https://api.example.com"
  auth_type: "bearer"
  rate_limit: 60
  
security:
  validate_inputs: true
  log_requests: true
  timeout: 30
Claude Desktop Integration

Generated servers automatically include configuration for Claude Desktop:

{
  "mcpServers": {
    "my-mcp-server": {
      "command": "python",
      "args": ["/path/to/my-mcp-server/server.py"],
      "env": {
        "API_KEY": "your-api-key",
        "DATABASE_URL": "your-db-connection"
      }
    }
  }
}
Development and Testing

Each generated project includes:

  • Unit tests with pytest
  • Syntax validation - all generated code is syntactically correct
  • MCP Inspector integration for testing
  • Comprehensive test suites for all tools and resources
  • Configuration validation and error handling
# Navigate to your generated project
cd my-mcp-server

# Install dependencies (if not done during creation)
pip install -r requirements.txt

# Validate the generated code syntax
python -c "import server; print('Server syntax OK')"

# Run the test suite
python -m pytest tests/ -v

# Test with MCP Inspector
npx @modelcontextprotocol/inspector python server.py

# Run the server directly
python server.py
Security Features
  • 🔒 Input validation and sanitization
  • 🛡️ Access controls and permission systems
  • 📝 Audit logging for all operations
  • 🔐 Secure credential management
  • 🚫 SQL injection prevention
  • 🌐 CORS and rate limiting
Quality Assurance

ModelContextKit generates high-quality, production-ready code:

  • Syntax Validation: All generated Python code passes syntax validation
  • MCP Compliance: Generated servers follow MCP protocol specifications
  • Type Safety: Proper type hints and schema definitions
  • Error Handling: Comprehensive exception handling and logging
  • Test Coverage: Generated test suites validate all functionality
  • Documentation: Auto-generated documentation and examples
CLI Reference
# Create new MCP server
modelctx create <project-name> --backend <type> [options]

# Interactive wizard
modelctx wizard

# List available backends
modelctx list

# Manage templates
modelctx templates [list|add|remove]

# Generate documentation
modelctx docs <project-path>

# Deploy server (if configured)
modelctx deploy <project-name>
Contributing

We welcome contributions! Please see our Contributing Guide for details.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

Made with ❤️ for the MCP community