mcp-rag-server-template

A starter MCP server template that exposes RAG-style retrieval tools using Python.
This project demonstrates how to connect document retrieval capabilities to an MCP-compatible client through a clean, language-agnostic architecture.

---

🚀 Features

MCP Server Capabilities

• MCP server runtime using Python
• Tool registration through FastMCP
• Local MCP client for testing tool calls
• Structured tool outputs suitable for downstream LLM use

Retrieval / RAG Capabilities

• Document loading from local files
• Text chunking
• Embedding generation using OpenAI
• Persistent vector storage with ChromaDB
• Retrieval of relevant chunks for a question
• Source tracking in results

Tooling

• list_sources_tool
• search_documents_tool
• retrieve_chunks_tool

---

🧠 Architecture

This repo separates concerns into layers:

• MCP layer → tool registration and runtime
• Handler layer → tool implementations
• RAG layer → document loading, chunking, embeddings, vector search
• Client layer → local test harness

Flow: User → MCP Client → MCP Server → Tools → RAG Layer → OpenAI + ChromaDB → Structured Results

---

🛠️ Exposed Tools

• list_sources_tool
• search_documents_tool
• retrieve_chunks_tool

---

📂 Project Structure

main.py
client.py

server/
rag/
data/
outputs/

---

⚡ Setup

pip install -r requirements.txt

Create .env: OPENAI_API_KEY=your_openai_api_key_here

---

▶️ Run Server

python main.py

---

🧪 Test Client

python client.py

---

⭐

Star, fork, and extend this MCP-based retrieval service.