🚀 Ultralytics MCP Server - AI-Powered Computer Vision Platform

Unified Development Platform for YOLO Models with N8N Integration

A comprehensive Model Context Protocol (MCP) server that seamlessly integrates Ultralytics YOLO models with N8N workflows, providing a complete AI-powered computer vision solution in a single command.

Docker CUDA Streamlit N8N

✨ Features

🎯 Core Capabilities

7 AI-Powered Tools for comprehensive YOLO operations
Real-time Object Detection with live inference
Model Training & Fine-tuning with custom datasets
Performance Analytics via TensorBoard integration
N8N Workflow Integration for automation

🖥️ User Interfaces

Streamlit Dashboard - Interactive web interface for model management
Jupyter Lab - Notebook environment for development
TensorBoard - Real-time training metrics and visualization
N8N Integration - Workflow automation and AI task orchestration

🔧 Technical Stack

CUDA 12.4.1 - GPU acceleration for training and inference
PyTorch - Deep learning framework with CUDA support
Ultralytics YOLO - State-of-the-art object detection models
Docker - Containerized deployment
Node.js MCP Server - Model Context Protocol implementation

🚀 Quick Start

Prerequisites

Docker Desktop with GPU support
NVIDIA drivers compatible with CUDA 12.4.1
Windows PowerShell or Linux/macOS terminal

One-Command Deployment

docker-compose up -d

That's it! The entire platform will be available at:

🌐 Streamlit UI: http://localhost:8501
📊 TensorBoard: http://localhost:6006
📓 Jupyter Lab: http://localhost:8888
🔗 MCP Server: http://localhost:8092

🎮 Available Services

Service	Port	Description	Status
Streamlit Dashboard	8501	Interactive YOLO model interface	✅ Ready
MCP Server	8092	N8N integration endpoint	✅ Ready
TensorBoard	6006	Training metrics visualization	✅ Ready
Jupyter Lab	8888	Development environment	✅ Ready

🛠️ MCP Tools Available

Our MCP server provides 7 specialized tools for AI workflows:

detect_objects - Real-time object detection in images
train_model - Custom YOLO model training
evaluate_model - Model performance assessment
predict_batch - Batch processing for multiple images
export_model - Model format conversion (ONNX, TensorRT, etc.)
benchmark_model - Performance benchmarking
analyze_dataset - Dataset statistics and validation

🔌 N8N Integration

Connect to N8N using our MCP server:

Server Endpoint: http://localhost:8092
Transport: Server-Sent Events (SSE)
Health Check: http://localhost:8092/health

Example N8N Workflow

{ "mcp_connection": { "transport": "sse", "endpoint": "http://localhost:8092/sse" } }

📁 Project Structure

ultralytics_mcp_server/ ├── 🐳 docker-compose.yml # Orchestration configuration ├── 🔧 Dockerfile.ultralytics # CUDA-enabled Ultralytics container ├── 🔧 Dockerfile.mcp-connector # Node.js MCP server container ├── 📦 src/ │ └── server.js # MCP server implementation ├── 🎨 main_dashboard.py # Streamlit main interface ├── 📄 pages/ # Streamlit multi-page app │ ├── train.py # Model training interface │ └── inference.py # Inference interface ├── ⚡ startup.sh # Container initialization script ├── 📋 .dockerignore # Build optimization └── 📖 README.md # This documentation

🔧 Configuration

Environment Variables

CUDA_VISIBLE_DEVICES - GPU device selection
STREAMLIT_PORT - Streamlit service port (default: 8501)
MCP_PORT - MCP server port (default: 8092)
TENSORBOARD_PORT - TensorBoard port (default: 6006)

Custom Configuration

Edit docker-compose.yml to customize:

Port mappings
Volume mounts
Environment variables
Resource limits

📊 Usage Examples

Object Detection via Streamlit

Navigate to http://localhost:8501
Upload an image or video
Select YOLO model (YOLOv8, YOLOv11)
Run inference and view results

Training Custom Models

Access Jupyter Lab at http://localhost:8888
Prepare your dataset in YOLO format
Use the training interface in Streamlit
Monitor progress in TensorBoard

N8N Automation

Create N8N workflow
Add MCP connector node
Configure endpoint: http://localhost:8092
Use available tools for automation

🔍 Monitoring & Debugging

Container Status

docker ps docker-compose logs ultralytics-container docker-compose logs mcp-connector-container

Health Checks

# MCP Server curl http://localhost:8092/health # Streamlit curl http://localhost:8501/_stcore/health # TensorBoard curl http://localhost:6006

🔄 Restart & Maintenance

Restart Services

docker-compose restart

Update & Rebuild

docker-compose down docker-compose up --build -d

Clean Reset

docker-compose down docker system prune -f docker-compose up --build -d

🎯 Performance Optimization

GPU Memory: Automatically managed by CUDA runtime
Batch Processing: Optimized for multiple image inference
Model Caching: Pre-loaded models for faster response
Multi-threading: Concurrent request handling

🚨 Troubleshooting

Common Issues

Container Restart Loop

# Check logs docker-compose logs ultralytics-container # Restart with rebuild docker-compose down docker-compose up --build -d

Streamlit Not Loading

# Verify container status docker ps # Check if files are copied correctly docker exec ultralytics-container ls -la /ultralytics/

GPU Not Detected

# Check NVIDIA drivers nvidia-smi # Verify CUDA in container docker exec ultralytics-container nvidia-smi

🔧 Development

Local Development Setup

Clone repository
Install dependencies: npm install (for MCP server)
Set up Python environment for Streamlit
Run services individually for debugging

Adding New MCP Tools

Edit src/server.js
Add tool definition in tools array
Implement handler in handleToolCall
Test with N8N integration

🤝 Contributing

Fork the repository
Create feature branch (git checkout -b feature/amazing-feature)
Commit changes (git commit -m 'Add amazing feature')
Push to branch (git push origin feature/amazing-feature)
Open Pull Request

📄 License

This project is licensed under the AGPL-3.0 License - see the Ultralytics License for details.

🙏 Acknowledgments

Ultralytics - For the amazing YOLO implementation
N8N - For the workflow automation platform
Streamlit - For the beautiful web interface framework
NVIDIA - For CUDA support and GPU acceleration

📞 Support

🐛 Issues: GitHub Issues
💬 Discussions: GitHub Discussions
📧 Contact: Create an issue for support

Made with ❤️ for the AI Community

🚀 Ready to revolutionize your computer vision workflows? Start with

This server cannot be installed

-

security - not tested

F

license - not found

-

quality - not tested

How are these scores calculated?

hybrid server

The server is able to function both locally and remotely, depending on the configuration or use case.

A Model Context Protocol compliant server that provides RESTful API access to Ultralytics YOLO operations for computer vision tasks including training, validation, prediction, export, tracking, and benchmarking.

Related MCP Servers

AI Vision Debug MCP Server
A
security
A
license
A
quality
A Model Context Protocol server that provides AI vision capabilities for analyzing UI screenshots, offering tools for screen analysis, file operations, and UI/UX report generation.
mcp-server-unitycatalog
ognis1205
-
security
A
license
-
quality
A Model Context Protocol server that provides access to Unity Catalog Functions, allowing AI assistants to list, get, create, and delete functions within Unity Catalog directly through a standardized interface.
Last updated -
15
MIT License
nUR MCP Server
nonead
A
security
F
license
A
quality
A Model Context Protocol server that enables natural language interactive control of Universal Robots collaborative robots, allowing users to control robot motion, monitor status, and execute programs through direct commands to large language models.
Last updated -
29
3
Luno MCP Server
amanasmuei
-
security
F
license
-
quality
A Model Context Protocol server that provides a standardized interface for AI models and applications to interact with the Luno cryptocurrency exchange API for trading operations.
Last updated -
2

View all related MCP servers