README.md

---
title: CV MCP Server
emoji: 💻
colorFrom: yellow
colorTo: blue
sdk: gradio
sdk_version: 6.0.1
app_file: app.py
pinned: false
license: mit
short_description: Real-time CV VLM MCP Server for MCP 1st Birthday Hackathon
tags:
- building-mcp-track-creative
---

# 🎥 CV MCP Server

A **Model Context Protocol (MCP) server** that provides **real-time scene analysis** for webcam images.  
This Space allows users to stream live video feeds and get detailed insights about the environment, objects, humans, and more.  

Check out the Hackathon details [here](https://huggingface.co/MCP-1st-Birthday).  
The social media post of this MCP Hackthon project on [Discord](https://discord.com/channels/879548962464493619/1439001549492719726/1443045145284051084) [Instagram](https://www.instagram.com/p/DRsw2KOADrB/) [Thread](https://www.threads.com/@oppa.ai_the.one.and.only/post/DRsxlNzAdCj?xmt=AQF0fVYU0qfeEUT4nDojv48yYZmjtK6tCrMx3sehnhVyOw).  

🎬 Watch a demo of the CV MCP Server analyzing a robot’s camera feed:  
[Demo Video](https://photos.app.goo.gl/guxui1EsdPNoL4mw7)  

GitHub repo of the cv robot python script:
https://github.com/OppaAI/CV_Robot_MCP

---

## 🌟 Key Features

- Real-time scene description  
- Human detection  
- Animal and object detection  
- Environment classification (indoor/outdoor)  
- Lighting condition analysis  
- Hazards identification  
- Optimized for **context window efficiency**  

---

## 🔧 How It Works

This Space leverages the MCP server to analyze images captured from your webcam:

1. Stream an image from your webcam.  
2. The image is sent to the **MCP server**.  
3. The server processes it using the **Model Context Protocol**.  
4. Outputs are returned and displayed in the UI, including:
   - General scene description  
   - Detected humans  
   - Detected animals and objects  
   - Environment type (indoor/outdoor)  
   - Lighting condition  
   - Hazards  

---

## ⚡ Demo

Compatible with **PC, mobile, and robots with cameras**.  

Stream images via your webcam or phone camera to receive real-time scene analysis.  

Watch a demo video of the CV MCP Server analyzing the video feed from my robot:  
[Demo Video](https://photos.app.goo.gl/guxui1EsdPNoL4mw7)  

---

## 🔑 Requirements

- A valid **Hugging Face API Token** is required.  
- If running locally, set your token as an environment variable:  

```bash
export HF_TOKEN=your_huggingface_token