🤖 Multi-Agent MCP - AI Knowledge Hub

A showcase repository demonstrating how to build multi-agent AI tools using Model Context Protocol (MCP), RAG, and LangChain. This project serves as both a playground and reference implementation for building intelligent, pluggable AI assistants.

🎯 Project Vision

Build a pluggable AI knowledge assistant where each "agent" is a microservice powered by MCP. Agents can be specialized for different tasks and collaborate through the MCP protocol, orchestrated by LangChain.

🏗️ Architecture

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   Web UI       │    │   LangChain     │    │   MCP Agents    │
│   (Next.js)    │◄──►│   Orchestrator  │◄──►│   (Microservices)│
└─────────────────┘    └─────────────────┘    └─────────────────┘
                              │
                              ▼
                       ┌─────────────────┐
                       │   RAG Pipeline  │
                       │   (Chroma/FAISS)│
                       └─────────────────┘

🚀 Core Features

• Pluggable Agent System: Add new agents without changing core code
• MCP Integration: Standardized communication between agents
• RAG Pipeline: Document processing and intelligent retrieval
• Multi-Agent Collaboration: Agents can hand off tasks to each other
• Extensible Connectors: Easy integration with external services

🎭 Agent Types

• 📚 Docs Agent: Answer questions from PDFs, Markdown, Confluence
• 💻 Code Agent: Explain, refactor, and generate code
• 🌐 Web Agent: Fetch real-time information via web search
• 💬 Chat Agent: General LLM conversation and task coordination

🛠️ Tech Stack

• Backend: Python, FastAPI, LangChain
• Frontend: Next.js, React, Tailwind CSS
• Database: ChromaDB (vector store), SQLite (metadata)
• Protocol: MCP (Model Context Protocol)
• AI: OpenAI, Anthropic (configurable)
• Deployment: Docker, Docker Compose

📦 Quick Start

Prerequisites

• Python 3.9+
• Node.js 18+
• Docker (optional)

Backend Setup

cd backend
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate
pip install -r requirements.txt
python main.py

Frontend Setup

cd frontend
npm install
npm run dev

🔧 Configuration

Create a .env file in the backend directory:

OPENAI_API_KEY=your_openai_key
ANTHROPIC_API_KEY=your_anthropic_key
CHROMA_HOST=localhost
CHROMA_PORT=8000

🤝 Contributing

This is a showcase project - feel free to fork and adapt for your own use cases!

📄 License

MIT License - see LICENSE file for details

---

Built with ❤️ using MCP, LangChain, and modern AI technologies