home / mcp / vybe virtual try-on mcp server

Vybe Virtual Try-On MCP Server

Provides virtual try-on capabilities by wrapping the Replicate API for MCP clients with health checks, image conversion utilities, and a 3-tool workflow.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "arnabgho-mcp-vybe": {
      "url": "https://your-service-name.onrender.com",
      "headers": {
        "HOST": "0.0.0.0",
        "PORT": "8080",
        "REPLICATE_API_TOKEN": "YOUR_TOKEN"
      }
    }
  }
}

Vybe Virtual Try-On MCP Server provides a Python-based MCP server that wraps the Replicate API to enable virtual garment try-on capabilities. It is designed for easy deployment and remote querying from MCP clients, allowing you to test connections, convert base64 images to usable data URIs, and perform live virtual try-ons with customizable options.

How to use

You connect your MCP client to the Vybe Virtual Try-On server to perform three main actions: test the connection, prepare images for the try-on workflow, and run the virtual try-on itself. Start by validating the health and readiness of the service, then convert any base64-encoded images you have into data URIs, and finally invoke the virtual_tryon tool with your model and garment images to preview the result. Ensure you have a valid Replicate API token to access the underlying model services.

Typical usage patterns include: testing connectivity to confirm the server responds promptly, transforming images for use in the try-on pipeline, and performing a full virtual try-on with a person image and a clothing item image. Each step is designed to integrate with MCP clients that support HTTP transport or local stdio configurations.

How to install

Prerequisites you need before installing are Python and an optional local tooling environment for development. If you prefer a local runtime with a package manager, you can choose between a Python-based setup or a UV-based workflow for faster development.

Install using UV (recommended) set up to run the server locally:

curl -LsSf https://astral.sh/uv/install.sh | sh
uv sync
cp .env.example .env
# Edit .env to add your Replicate API token

Install using pip (alternative):

pip install -r requirements.txt
cp .env.example .env
# Edit .env to add your Replicate API token

Run the server locally using UV or regular Python:

# With uv (recommended)
uv run python server.py

# Or with regular python
python server.py

Additional sections

Configuration, deployment, and testing details are covered below to help you run the server in production and verify a healthy deployment. The service exposes endpoints for health checks and status, and you can configure a remote MCP client to connect over HTTP.

Configuration and deployment notes

The server supports a remote MCP connection via an HTTP URL. You can point an MCP client to the deployed endpoint and specify the transport method as HTTP.

Examples of deployment and testing steps are provided to help you validate a running instance and confirm MCP connectivity.

Testing and deployment helpers

Health checks and basic connectivity tests are available to verify the service is healthy and responsive after deployment.

If you encounter timeouts during long-running Replicate operations, you can adjust timeout and polling settings in the server configuration.

Available tools

test_connection

Test the connection to the MCP server and verify timeout handling and basic responsiveness.

base64_to_url

Convert a base64 encoded image into a data URI suitable for use as model_image or garment_image in virtual_tryon. Takes input base64 string and image_type with a default of png.

virtual_tryon

Perform a live virtual try-on by supplying a person/model image and a garment image, with optional customization parameters for the output.