README.md•1.55 kB
# 🧠 MCP AI Server — Modular Context Protocol for Intelligent Search
Welcome to the **MCP AI Server**, a powerful and modular tool that uses **RAG-based retrieval**, **Pinecone vector storage**, and **MCP** (Model Context Protocol) to create intelligent assistants capable of answering domain-specific questions from your own knowledge base.
---
## 🚀 Features
✅ Local MCP server with FastAPI + Claude/ChatGPT integration
✅ Embedding using `intfloat/multilingual-e5-large` (via SentenceTransformer)
✅ Fast vector search with Pinecone
✅ Documented `tools` exposed to clients like **Claude** and **Cursor IDE**
✅ Secure `.env` usage for managing API keys
✅ Clean, extensible architecture
---
## 🔧 Setup Instructions
### 1. Clone the Repo
```bash
git clone git@github.com:MeetRathodNitsan/MCP1.git
cd MCP1
```
### 2. Create a Virtual Environment
```bash
python -m venv .venv
# Windows
.venv\Scripts\activate
# macOS/Linux
source .venv/bin/activate
```
### 3. Install Dependencies
```bash
pip install -r requirements.txt
```
### 4. Configure Environment Variables
```bash
OPENAI_API_KEY=your-api-key...
PINECONE_API_KEY=...
PINECONE_ENVIRONMENT=your-env
```
### 5. How to use it
```bash
uv --directory F:/Project run main.py
```