MCPGex

MCP server for finding, testing and refining regex patterns

MCPGex is an MCP server that allows LLMs to test and validate regex patterns against test cases. It provides a systematic way to develop regex patterns by defining or generating expected outcomes and iteratively testing patterns until all requirements are satisfied.

Index

• How it works
• Installation
• Usage
  • Running the Server
  • Configuration
  • Available Tools
• Benefits
• Requirements
• License

How it works

1. Define the goal: You provide what the goal regex pattern should return. The LLM will generate test cases for you.
2. Test patterns: The LLM can test different regex patterns against all defined test cases to see which ones pass or fail.
3. Iterate: Based on the results, the LLM can refine the regex pattern until all test cases pass.
4. Validate: Once all tests pass, you have a regex pattern that works for your specific use cases.

Installation

Go ahead and install through pip:

pip3 install mcpgex

Usage

Running the Server

If you want to start the MCP server:

mcpgex

Configuration

You can also add a configuration. For example, for Claude Desktop, you can have:

{
  "mcpServers": {
    "mcpgex": {
      "command": "python3",
      "args": ["-m", "mcpgex"]
    }
  }
}

Or, for other applications, such as Zed:

"context_servers": {
  "mcpgex": {
    "command": {
      "path": "python3",
      "args": ["-m", "mcpgex"]
    },
    "settings": {}
  }
}

Then, you will be able to use the server in these tools without having to run the python script manually!

{
  "input_string": "Contact me at john@example.com for details", 
  "expected_matches": ["john@example.com"],
  "description": "Basic email extraction"
}

{
  "pattern": "[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}",
  "flags": "i"
}

Benefits

• Comprehensive testing: Ensure patterns work across various use cases
• Iterative improvement: Easy to test and refine patterns
• Documentation: Test cases serve as examples and documentation
• Confidence: Know your regex works before deploying it
• Fully Automated: Give it instructions, let it do the rest

Requirements (installed automatically through pip3)

• Python 3.8+
• MCP library (pip3 install mcp)

License

This project is open source under the MIT license. Feel free to use and modify as needed.