CodexAid - MCP Self-Learning Service

A powerful service for building, managing, and deploying self-learning and self-generating MCP (Model Context Protocol) agents.

Overview

CodexAid is a framework that allows you to create intelligent MCP agents that can learn from tasks and generate new skills on their own. The system follows a design-develop-deploy pipeline that turns natural language prompts into functional MCP servers with custom tools.

Features

• Self-Learning: Create agents that can learn from user interactions
• Self-Generation: Generate new MCP servers based on natural language prompts
• Tool Generation: Automatically create and deploy new tools based on requirements
• FastAPI Integration: Expose MCP functionality through an HTTP API
• Design-Develop Pipeline: Structured approach to converting tasks into working code

Architecture

The system consists of several key components:

• Designer: Analyzes tasks and creates specifications for development
• Developer: Implements tools based on specifications
• MCP Builder: Constructs MCP servers from developed tools
• Tool Server: FastAPI server that hosts and manages MCP tools

Getting Started

Prerequisites

• Python 3.12 or higher
• Dependencies listed in pyproject.toml

Installation

# Clone the repository
git clone <repository-url>
cd CodexAid

# Create and activate a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -e .

Usage

1. Start the Tool Server:

python mcp_service/tool_server.py

2. Create a new MCP from a prompt:

from scr.work import mcp_create

# Generate a new MCP server from a natural language prompt
mcp_str = mcp_create("Create a tool that summarizes web content")

# Save the generated server
with open("my_new_mcp.py", "w") as f:
    f.write(mcp_str)

3. Use the HTTP API to generate new MCPs:

curl -X POST "http://localhost:8848/new_mcp" \
     -H "Content-Type: application/json" \
     -d '{"prompt": "Create a tool that searches for information"}'

API Endpoints

• GET /health - Check server health
• GET /reset - Reset the MCP manager
• GET /get_tool - Get available tools
• GET /get_tool/{agent_name} - Get tools for a specific agent
• POST /call_tool/{tool_name} - Call a specific tool
• POST /new_mcp - Generate a new MCP from a prompt

Configuration

Edit the config.py file to configure:

• LLM API endpoints and keys
• MCP service paths
• Server port

Examples

Creating a Learning Timer

from mcp.server.fastmcp import FastMCP
mcp = FastMCP("LearningTimer")

@mcp.tool(description='Start a learning timer when detected')
async def start_learning_timer():
    # Implementation details
    pass

if __name__ == '__main__':
    mcp.run(transport='stdio')

License

See the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.