A collection of experiments and implementations with Model Context Protocol (MCP) servers and everything in between.
This repository serves as a playground for exploring the Model Context Protocol (MCP), which enables AI assistants to securely connect to external data sources and tools. Our experiments range from simple data processing servers to more complex integrations.
The Model Context Protocol is an open standard that allows AI assistants to work with external tools and data sources in a secure, standardized way. MCP servers can provide:
- Tools for data processing and analysis
- Access to databases and file systems
- Integration with external APIs
- Custom business logic and workflows
mcp-recipes/
├── hello-world/ # Basic MCP server example
│ ├── data/ # Sample data files (CSV, Parquet)
│ ├── tools/ # Tool implementations
│ │ ├── csv_tools.py # CSV processing tools
│ │ └── parquet_tools.py # Parquet processing tools
│ ├── utils/ # Utility functions
│ ├── server.py # Main MCP server implementation
│ ├── main.py # Server runner
│ └── README.md # Detailed documentation
└── README.md # This file
- Python 3.8+
- uv (Python package manager)
-
Navigate to an example directory:
cd hello-world -
Install dependencies:
uv sync
-
Run the MCP server:
uv run python main.py
The hello-world directory contains a comprehensive example MCP server that demonstrates:
- CSV and Parquet file processing
- Data analysis tools
- File system operations
- Structured data handling
This server provides tools for reading, analyzing, and manipulating tabular data in various formats.
Feel free to add your own MCP server experiments to this repository. Each experiment should:
- Be in its own directory
- Include a README with setup and usage instructions
- Follow the MCP specification
- Include example data or test cases where applicable
This project is open source and available under the MIT License.