Agent Construct
"We can load anything, from clothing to equipment, weapons, training simulations, anything we need." - The Matrix (1999)
Agent Construct is a Model Context Protocol (MCP) server implementation that standardizes how AI applications access tools and context. Just as the Construct in The Matrix provided operators with instant access to any equipment they needed, Agent Construct provides a standardized interface for AI models to access tools and 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 for AI applications. It provides a robust and scalable way to expose capabilities to AI models through a standardized protocol. It also provides a simplified configuration and tool structure to make adding new capabilities a breeze! An example tool for searching the web with Gemini is included.
Core Features
MCP Protocol Implementation
- Full MCP Compliance : Complete implementation of the Model Context Protocol specification
- Tool Discovery : Dynamic tool registration and discovery mechanism
- Standardized Communication : Implements MCP's communication patterns for tool interaction
Server Architecture
- FastAPI Backend : High-performance asynchronous server implementation
- Event Streaming : Real-time updates via Server-Sent Events (SSE)
- Modular Design : Clean separation between core protocol handling and tool implementations
- Handler System : Extensible request handler architecture for different MCP operations
- Tool-Based Rate Limiting : Let the server handle your configurable per-tool rate limiting.
Development Features
- Tool Decorator System : Simple way to expose new tools via MCP
- Logging & Monitoring: Comprehensive logging system for debugging and monitoring
- Configuration Management : Environment-based configuration with secure defaults
- Testing Framework : Extensive test suite for protocol compliance
- Agent Framework Friendly : Included implementation examples for custom clients or frameworks like smolagents.
Getting Started
Prerequisites
- Python 3.8 or higher
- pip package manager
Installation
Clone the repository:
git clone https://github.com/yourusername/agent-construct.git
cd agent-construct
Install dependencies:
pip install -r requirements.txt
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 for production
Run the server:
python -m mcp_server
Core Architecture
mcp_server/
├── core/ # Core MCP protocol implementation
│ ├── server.py # Main server implementation
│ ├── protocol.py # MCP protocol handlers
│ └── 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 utilities
│ └── 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 scoping and isolation
- Real-time context updates
- Context persistence options
Communication Patterns
- Synchronous request/response
- Server-sent events for updates
- Streaming responses
- Error handling and recovery
Future Enhancements
Protocol Extensions
Security
Performance
Development
Contributing
Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.
License
This project is licensed under the MIT License - see the LICENSE file for details.
Acknowledgements
- Model Context Protocol for the protocol specification
- FastAPI for the excellent web framework
- The open-source community for various tools and libraries used in this project
