Agent Construct

🚀 Agent Construct

Agent Construct is a Model Context Protocol (MCP) server implementation that standardizes the way AI applications access tools and context, offering a powerful and scalable solution for AI models to expose capabilities.

Agent Construct is an implementation of the Model Context Protocol (MCP) server, which standardizes the way AI applications access tools and context. Just as the Construct in The Matrix provided operators with an interface to instantly access the equipment they needed, Agent Construct provides AI models with a standardized interface to access tools and service data through the MCP specification.

Built on the Model Context Protocol specification, it acts as a central hub that manages tool discovery, execution, and context management, offering a powerful and scalable way for AI applications to expose capabilities to models through the MCP specification. It also provides a simplified configuration and tool structure, making it a breeze to add new features! It includes an example tool for web searching using Gemini.

✨ Features

MCP Protocol Implementation

• Full MCP Compatibility: Fully implements the Model Context Protocol specification.
• Tool Discovery: Dynamic tool registration and discovery mechanism.
• Standardized Communication: Implements the MCP communication mode for tool interaction.

Server Architecture

• FastAPI Backend: High-performance asynchronous server implementation.
• Event Streaming: Real-time updates provided via Server-Sent Events (SSE).
• Modular Design: Clear separation between core protocol handling and tool implementation.
• Processor System: Scalable request processor architecture to support different MCP operations.
• Tool-based Rate Limiting: Allows the server to handle configuration-based tool rate limiting.

Development Features

• Tool Decorator System: A simple way to expose new tools through MCP.
• Logging and Monitoring: Comprehensive logging system for debugging and monitoring.
• Configuration Management: Environment-based configuration with secure defaults.
• Testing Framework: Comprehensive test suite to ensure protocol compliance.
• Friendly Agent Framework: Includes implementation examples for custom clients or frameworks like smolagents.

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• pip package manager

📦 Installation

1. Clone the repository:

   git clone https://github.com/yourusername/agent-construct.git
   cd agent-construct
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Set up environment variables: Create a .env file in the root directory with the following variables:

   # Server configuration
   SERVER_HOST=localhost
   SERVER_PORT=8000
   
   # MCP protocol settings
   MCP_VERSION=1.0
   TOOL_DISCOVERY_ENABLED=true
   
   # Security settings
   ENABLE_AUTH=false  # Enable in production
   

4. Run the server:

   python -m mcp_server
   

📚 Documentation

Core Architecture

mcp_server/
├── core/               # Core MCP protocol implementation
│   ├── server.py      # Main server implementation
│   ├── protocol.py    # MCP protocol processor
│   └── context.py     # Context management
├── handlers/          # MCP operation handlers
│   ├── discovery.py   # Tool discovery
│   ├── execution.py   # Tool execution
│   └── context.py     # Context operations
├── utils/            # Utility functions
│   ├── logging.py    # Logging configuration
│   ├── security.py   # Security tools
│   └── config.py     # Configuration management
└── __main__.py       # Server entry point

MCP Protocol Features

Tool Discovery

• Dynamic tool registration system
• Tool capability advertisement
• Version management
• Tool metadata and documentation

Context Management

• Efficient context storage and retrieval
• Context loading/unloading
• Context sharing and isolation

Other Features

• Security: Achieves resource isolation between different models through configuration, ensuring that each model can only access the tools and services it is permitted to use.
• Scalability: Supports adding new contexts and tools to meet the needs of different models.

📄 License

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

Acknowledgments

• Model Context Protocol for providing the protocol specification.
• FastAPI for the excellent web framework.
• Various tools and libraries from the open-source community used to build this project.