🤖 mcp-agent-orchestrator: Local Agentic AI via Model Context Protocol

mcp-agent-orchestrator is a production-ready starter template for building local-first Agentic AI systems. It demonstrates how to connect a FastMCP (Python) Server to a LangGraph Orchestrator using Ollama as the local LLM host.

🚀 Key Features

• Standardized MCP Integration: Uses the Model Context Protocol to decouple tool logic from LLM logic.
• Stateful Orchestration: Built with LangGraph for robust, multi-turn agent conversations and error handling.
• Privacy-First: 100% local execution using Ollama (Llama 3.1/Mistral)—no API keys required.
• Modern UI: A clean Streamlit interface for real-time interaction with your MCP tools.
• FastMCP Framework: Simplified Python tool definition with automatic schema generation.

🏗️ Architecture Overview

1. MCP Host (Ollama): Provides the reasoning engine (LLM).
2. MCP Server (FastMCP): Defines Python tools (e.g., SQLite, File System, Web Search).
3. MCP Client: Manages the stdio transport and subprocess communication.
4. Orchestrator (LangGraph): A stateful ReAct agent that decides when to call MCP tools.
5. Interface (Streamlit): The user-facing web application.

🛠️ Quick Start

Prerequisites

• Install Ollama and run ollama run llama3.1.
• Python 3.10 or higher.

Installation

1. Clone the repository:

   git clone https://github.com/NxtGenCodeBase/mcp-agent-orchestrator.git
   cd mcp-agent-orchestrator