MCP-Swarm Framework
Distributed AI Agent Orchestration via Model Context Protocol
🚀 The Innovation
While MCP has gained traction with early adopters, most implementations focus on single-agent scenarios. MCP-Swarm is the first framework to enable MCP-based multi-agent orchestration, where agents communicate through MCP servers rather than traditional message passing.
✨ Key Features
- Protocol-First Design: All communication through standardized MCP
- Agent Autonomy: Each agent operates as an independent MCP server
- Dynamic Discovery: Agents discover each other via MCP Registry
- State Synchronization: Distributed state management across agents
- Production-Ready: Built with observability, resilience, and scalability in mind
📋 Requirements
- Python 3.11+
- UV package manager
🛠️ Installation```bash
Clone the repository
git clone https://github.com/rinaldofesta/mcp-swarm.git
cd mcp-swarm
Install dependencies
uv sync
Activate virtual environment (UV handles this automatically)
uv run python --version
## 🎯 Quick Start```python
from mcp_swarm.core.agent import MCPAgent
# Create an agent
agent = MCPAgent(name="my-agent", role="processor")
# Initialize and start
await agent.initialize()
await agent.start()
📚 Documentation
- Architecture Guide
- API Reference
- Tutorials
- Examples
🏗️ Project Structure```
mcp-swarm/
├── src/mcp_swarm/ # Core framework code
├── tests/ # Test suite
├── examples/ # Example implementations
├── docs/ # Documentation
└── scripts/ # Utility scripts
```
🤝 Contributing
We welcome contributions! Please see our Contributing Guide for details.
📄 License
This project is licensed under the MIT License - see the LICENSE file for details.
🙏 Acknowledgments
Built with FastMCP and following the Model Context Protocol specification.
Status: 🚧 Early Development - We're building the future of AI agent orchestration!
