RAGwithMCP

A local project that combines Retrieval-Augmented Generation (RAG) with the Model Context Protocol (MCP) so you can:

1. Upload documents (PDF, DOCX, TXT, MD, CSV) via a web GUI.
2. Chat with your documents using an LLM powered by OpenRouter.
3. Connect any MCP-compatible LLM (Claude Desktop, Cursor, Cline, etc.) to your personal knowledge base via the included MCP server.

---

Architecture

┌─────────────────────────────────────────────┐
│               Gradio Web GUI                │
│  Upload ▸ Manage ▸ Chat ▸ MCP Server Info   │
└────────────────────┬────────────────────────┘
                     │
        ┌────────────▼────────────┐
        │   RAG Pipeline          │
        │  document_processor.py  │  ← text extraction + chunking
        │  vector_store.py        │  ← ChromaDB (local, persistent)
        └────────────┬────────────┘
                     │
        ┌────────────▼────────────┐        ┌──────────────────────────┐
        │   llm/openrouter.py     │        │   mcp_server.py          │
        │   OpenRouter API        │        │   MCP tools (stdio)      │
        │   openrouter/auto       │        │   search / list / add /  │
        └─────────────────────────┘        │   delete documents       │
                                           └──────────────────────────┘

---

Quick Start

1. Install dependencies

pip install -r requirements.txt

2. Configure your API key

cp .env.example .env
# Edit .env and set your OPEN_ROUTER_API_KEY

3. Start the web GUI

python main.py
# Open http://localhost:7860 in your browser

4. (Optional) Start the MCP server

In a separate terminal:

python mcp_server.py

---

Using the Web GUI

Tab	What you can do
📤 Upload Documents	Upload PDF / DOCX / TXT / MD / CSV files; configure chunk size & overlap
🗂️ Manage Documents	View all stored documents and delete individual ones
💬 Chat	Ask questions answered with RAG context from your uploaded files
🔌 MCP Server	Instructions for connecting the MCP server to your LLM

---

MCP Server

The MCP server exposes your knowledge base as tools via the stdio transport.

Available tools

Tool	Description
search_knowledge_base	Find relevant chunks by query
list_documents	List all uploaded documents
add_document_text	Add raw text to the knowledge base
delete_document	Remove a document

Connect to Claude Desktop

Add this to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "rag-knowledge-base": {
      "command": "python",
      "args": ["/absolute/path/to/RAGwithMCP/mcp_server.py"]
    }
  }
}

Connect with the MCP CLI

mcp dev mcp_server.py

---

Environment Variables

Variable	Default	Description
OPEN_ROUTER_API_KEY	(required)	Your OpenRouter API key
OPENROUTER_MODEL	openrouter/auto	Model to use (e.g. openrouter/auto)
CHROMA_DB_PATH	./chroma_db	Path to the ChromaDB persistent store
PORT	7860	Port for the Gradio web interface

---

Project Structure

RAGwithMCP/
├── main.py                  # Entry point (Gradio GUI)
├── mcp_server.py            # MCP server (stdio transport)
├── requirements.txt
├── .env.example
├── rag/
│   ├── document_processor.py  # File reading + text chunking
│   └── vector_store.py        # ChromaDB wrapper
├── llm/
│   └── openrouter.py          # OpenRouter API client
└── gui/
    └── app.py                 # Gradio interface