generated: '2026-08-28'
method: searched
status: published
source: https://docs.ximilar.com/mcp
provider: Ximilar
providerId: ximilar
description: >-
  Ximilar publishes a first-party MCP server for its Computer Vision Platform. It is a
  Python FastMCP server distributed inside the official ximilar-client repository
  (ximilar/server/server.py) and run locally over stdio by the user's MCP client; there
  is no hosted/remote MCP endpoint. The tool list below was read from the provider's MCP
  documentation page AND verified against the server source, which is why fashion_region_id
  appears here even though the docs table omits it.
deployment:
  mode: local-stdio
  endpoint: null
  install: 'fastmcp install MCP_CLIENT_NAME PATH_TO_XIMILAR_CLIENT/ximilar/server/server.py --with fastmcp --with requests --env XIMILAR_API_KEY=API_KEY'
  package: https://gitlab.com/ximilar-public/ximilar-client
  auth: api-key
  verified: searched
  note: >-
    Verified by reading https://docs.ximilar.com/mcp (HTTP 200) and the server source at
    https://gitlab.com/ximilar-public/ximilar-client/-/raw/master/ximilar/server/server.py
    (HTTP 200, FastMCP("ximilar-mcp-server")). No hosted endpoint is advertised anywhere in
    the docs, and no mcp.ximilar.com host resolves; nothing was guessed.
server:
  name: ximilar-mcp-server
  transport: stdio
  framework: fastmcp
  runtime: python >= 3.10
  source_file: https://gitlab.com/ximilar-public/ximilar-client/-/raw/master/ximilar/server/server.py
  auth:
    type: api-key
    env: XIMILAR_API_KEY
    header: 'Authorization: Token <API_KEY>'
client_config:
  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"}
        }
      }
    }
tools:
  - name: grade_card
    category: Card Grading
    description: Grade collectible cards such as TCGs and sports cards (corners, edges, surface, centering).
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/card-grader/v2/
    docs: https://docs.ximilar.com/collectibles/card-grading
  - name: tcg_card_id
    category: Collectibles
    description: Identify trading card game (TCG) cards from an image URL or file path.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/collectibles/v2/tcg_id
    docs: https://docs.ximilar.com/collectibles/recognition
  - name: sports_card_id
    category: Collectibles
    description: Identify sports cards from an image URL or file path.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/collectibles/v2/sport_id
    docs: https://docs.ximilar.com/collectibles/recognition
  - name: comic_book_and_manga_id
    category: Collectibles
    description: Identify comic books and manga from an image.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/collectibles/v2/comics_id
    docs: https://docs.ximilar.com/collectibles/recognition
  - name: tcg_price_data_search
    category: Collectibles
    description: Search prices and listings for TCG cards using a text query (name + set + card number).
    inputs: [query]
    rest_endpoint: https://api.ximilar.com/collectibles/text/v2/tcg/pricing
    docs: https://docs.ximilar.com/collectibles/text-search
  - name: visual_search
    category: Similarity
    description: Find visually similar images with KNN search inside one of your collections.
    inputs: [img_path, collection_id, k]
    rest_endpoint: https://api.ximilar.com/similarity/photos/v2/visualKNN
    docs: https://docs.ximilar.com/visual-search/overview
  - name: fashion_detect_tags
    category: Fashion Tagging
    description: Detect fashion objects and their tags; optionally tag every detected object rather than the largest.
    inputs: [img_path, detect_tags_for_all_objs]
    rest_endpoint: https://api.ximilar.com/tagging/fashion/v2/detect_tags
    rest_endpoint_all: https://api.ximilar.com/tagging/fashion/v2/detect_tags_all
    docs: https://docs.ximilar.com/tagging/fashion
  - name: fashion_meta_tags
    category: Fashion Tagging
    description: Return fashion meta tags — Background, Scene, View and Person part.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/tagging/fashion/v2/meta
    docs: https://docs.ximilar.com/tagging/fashion
  - name: fashion_region_id
    category: Fashion Tagging
    description: Detect the fashion region of an image (e.g. Indian, generic).
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/tagging/fashion/v2/region
    docs: https://docs.ximilar.com/tagging/fashion
    note: Present in the server source but not listed in the docs endpoint table.
  - name: fashion_detect
    category: Fashion Tagging
    description: Detect fashion objects and their locations only, without tagging.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/tagging/fashion/v2/detect
    docs: https://docs.ximilar.com/tagging/fashion
  - name: fashion_tags
    category: Fashion Tagging
    description: Return fashion tags and attributes only, without detection.
    inputs: [img_path]
    rest_endpoint: https://api.ximilar.com/tagging/fashion/v2/tags
    docs: https://docs.ximilar.com/tagging/fashion
coverage:
  tools_published: 11
  tools_in_docs_table: 10
  rest_surface_covered: >-
    The MCP server exposes Collectibles, Card Grading, Fashion Tagging and Visual Search
    only. The documented REST surface is much larger — Computer Vision Platform training
    (recognition, detection, segmentation), VLM, Home Decor, Photo Tagging, OCR, Upscaler,
    Background Removal, Product Description and the async request API have no MCP tool.
  crosswalk: >-
    No tool-crosswalk artifact is emitted: Ximilar publishes no OpenAPI, so there are no
    operationIds to bind these tools to. The backing REST endpoint URL is recorded inline
    on each tool instead.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com