Vybe Virtual Try-On MCP Server
A FastMCP server that wraps the Replicate API for virtual try-on functionality, ready for deployment to Render.
Setup
Using UV (recommended)
- Install uv if you haven't already:
- Install dependencies:
- Set up environment variables:
Edit .env
and add your Replicate API token.
Using pip (alternative)
- Install dependencies:
- Set up environment variables:
Edit .env
and add your Replicate API token.
Local Testing
Run the server locally:
The server will start on the default FastMCP port and expose the virtual_tryon
tool.
MCP Client Configuration
Add to your Claude Desktop or other MCP client configuration:
Usage
The server exposes three tools:
test_connection
Tests the connection and shows timeout configuration.
base64_to_url
Converts base64 encoded images to data URIs for use with virtual_tryon:
base64_image
: Base64 encoded image string (with or without data prefix)image_type
: Image type (png, jpg, jpeg, gif, webp) - default: png
Returns a data URI that can be used as model_image
or garment_image
in virtual_tryon.
virtual_tryon
Performs virtual try-on with these parameters:
model_image
: URL or data URI of the person/model imagegarment_image
: URL or data URI of the clothing item to try on- Various optional parameters for customization
Timeout Configuration
The server is configured with extended timeouts to handle long-running Replicate operations:
- MCP request timeout: 600 seconds (10 minutes)
- Replicate polling interval: 5 seconds
- Replicate timeout: 600 seconds (10 minutes)
If you still experience timeouts, you can adjust these in the server.py file.
Deployment to Render
Quick Deploy with render.yaml
- Push your code to a GitHub repository
- Connect the repository to Render
- Render will automatically detect the
render.yaml
configuration - Add your
REPLICATE_API_TOKEN
in the Render dashboard under Environment Variables - Deploy!
The service includes health check endpoints:
/health
- Health check endpoint for Render monitoring/
- Root endpoint with service status
Manual Setup
If you prefer manual configuration:
- Create a new Web Service on Render
- Connect your GitHub repository
- Configure the service:
- Build Command:
curl -LsSf https://astral.sh/uv/install.sh | sh && source $HOME/.cargo/env && uv sync --frozen --no-dev
- Start Command:
uv run python server.py
- Health Check Path:
/health
- Build Command:
- Add environment variables:
REPLICATE_API_TOKEN
: Your Replicate API token (required)PORT
: (leave empty, Render will auto-assign)HOST
: (leave empty, defaults to 0.0.0.0)
Using the Remote MCP Server
Once deployed, you can access the server at:
- Health Check:
https://your-service-name.onrender.com/health
- Root:
https://your-service-name.onrender.com/
- MCP Protocol: Use the base URL for MCP client connections
For MCP clients that support HTTP transport:
Replace your-service-name
with your actual Render service URL.
Testing the Deployment
Use the included test script to verify your deployment is working:
This will test:
- Health check endpoint (
/health
) - Root endpoint (
/
) - Basic MCP server connectivity
Example output:
Docker Deployment (Alternative)
A Dockerfile is also included if you prefer containerized deployment. Render will automatically detect and use it if present.
This server cannot be installed
hybrid server
The server is able to function both locally and remotely, depending on the configuration or use case.
A FastMCP server that provides virtual try-on functionality through the Replicate API, allowing users to visualize how clothing items would look on models.
Related MCP Servers
- AsecurityFlicenseAqualityA FastMCP server implementation that facilitates resource-based access to AI model inference, focusing on image generation through the Replicate API, with features like real-time updates, webhook integration, and secure API key management.Last updated -1815Python
- -securityAlicense-qualityA FastMCP server implementation that provides a standardized interface for accessing AI models hosted on Replicate's API, currently supporting image generation with customizable parameters.Last updated -3PythonMIT License
- AsecurityAlicenseAqualityAn MCP server that allows users to generate images using Replicate's Stable Diffusion model and save them to the local filesystem.Last updated -3PythonMIT License
- -securityAlicense-qualityA TypeScript-based MCP server that implements virtual try-on capabilities using the HeyBeauty API, allowing users to visualize how clothes would look on them through Claude.Last updated -40419JavaScriptMIT License