Ximilar - MCP Server
Before you can make requests to the Ximilar API, you will need to grab your API key from your settings (via Ximilar APP). You find it under Settings » Token. Be sure to have also proper plan activated, most of the endpoints are behind the Business plan or higher.
Overview
This page describes how to use the Ximilar MCP Server to access the functionality of the Ximilar Computer Vision Platform within your favorite LLM client.
The MCP (Multi-Channel Protocol) Server is a Python-based server that allows you to interact with the Ximilar API using a simple and consistent interface. It provides endpoints for Card Identification, Grading, Fashion Tagging, and more. All endpoints are available in MCP server are here.
Installation
To use MCP Server, you need to:
- Download or clone the Ximilar Client files from here (no need to install it) or directly the MCP Server.
git clone https://github.com/ximilar/ximilar-client.git
-
Install python 3.10 or higher. See instructions on realpython.com.
-
Install
fastmcp
Python package by running the following command in your terminal via pip:
pip install fastmcp
- Install the
uv
package manager from here. - Install the MCP Server by running this command in your terminal:
fastmcp install MCP_CLIENT_NAME PATH_TO_XIMILAR_CLIENT/ximilar/server/server.py --with fastmcp --with requests --env XIMILAR_API_KEY=API_KEY
- Replace
API_KEY
with your actual Ximilar API key. You can find it in Ximilar APP. - Replace
MCP_CLIENT_NAME
with the name of the MCP Client you want to use (e.g.,clade-desktop
,claude-code
) - Replace
PATH_TO_XIMILAR_CLIENT
with the path to the Ximilar Client files you downloaded in step 1. Do NOT delete the downloaded files afterwards - Note: this approach might not work with all MCP Clients, for more information, please refer to the FastMCP documentation (mainly the
Integrations
section)
- Start your MCP Client.
Python MCP Clients
Some examples in MCP clients, implemented in python are available here:
https://gitlab.com/ximilar-public/ximilar-client/-/tree/master/ximilar/server/tests
Claude Desktop
You can also use the following configuration for Claude Desktop:
{
"mcpServers": {
"ximilar-mcp": {
"command": "uv",
"args": [
"run",
"--with","fastmcp",
"--with","requests",
"fastmcp", "run", "PATH_TO_XIMILAR_CLIENT/ximilar/server/server.py"
],
"env": {
"XIMILAR_API_KEY": "YOUR_API_KEY"
}
}
}
}
Available Endpoints
The following endpoints are currently available to the MCP Server:
Service | MCP Endpoint | Description | Link |
---|---|---|---|
Card Grading | grade_card | Grade collectible cards for condition assessment | Link |
Collectibles | tcg_card_id | Identify Trading Card Game (TCG) cards | Link |
Collectibles | sports_card_id | Identify Sports cards | Link |
Collectibles | comic_book_and_manga_id | Identify Comic Book and Manga cards | Link |
Collectibles | tcg_price_data_search | Get pricing information for TCG cards | Link |
Similarity | visual_search | Find visually similar images using KNN search | Link |
Fashion Tagging | fashion_detect_tags | Detect fashion objects and tags for clothing items | Link |
Fashion Tagging | fashion_meta_tags | Get metadata about fashion tagging images | Link |
Fashion Tagging | fashion_detect | Detect fashion items and their locations | Link |
Fashion Tagging | fashion_tags | Recognize fashion tags and attributes for fashion items | Link |