What ImgVX AI MCP server does
The ImgVX AI MCP server gives AI clients structured access to ImgVX AI’s browser-based image-tool catalog. It helps an agent identify a suitable tool for a described task, browse the available tools, and check the service’s pricing. The server does not edit, upload, or download images on the client’s behalf. Instead, it returns links that users can open in their browser.
The catalog covers tasks including background removal, image upscaling, object erasure, photo enhancement, colorization, compression to a target file size, resizing, cropping, and format conversion. Supported formats mentioned in the listing include HEIC, PNG, JPG, WebP, and PDF. The most-used tools have pages in 40 languages, and the matching tool can be requested in a specified language.
How it works
ImgVX AI is provided as a hosted MCP endpoint at https://imgvx.com/mcp. It uses Streamable HTTP and requires no authentication. The server exposes three read-only tools:
find_image_tool accepts a short description of the desired image task and an optional language code. It returns up to five matching tools, including descriptions, prices, and links.
list_image_tools optionally filters the catalog by category and returns each tool’s short description and link.
get_pricing returns information about free features, Pro access, and credit packs.
The available categories are ai, compress-resize, adjust, effects, edit, convert, inspect, and create. Most tools are free, while HD downloads use credits. Image work takes place in the user’s browser rather than on the MCP server. The server receives only the tool inputs sent by the connected application, does not store them or associate them with a person, and sets no cookies according to the provided documentation.
Setup and configuration
No local package installation or environment variable is required. Add the endpoint as a custom MCP connection using the client’s HTTP configuration. Claude can use a custom connector or the Claude Code command with the HTTP transport. Cursor, VS Code, and Windsurf each support a JSON configuration pointing to the same URL. Zapier, Make, and n8n can connect through their MCP client features with Streamable HTTP and no authentication.
A direct request can also call the MCP endpoint using JSON-RPC. The documented example invokes find_image_tool with a task and Spanish language code. This makes the ImgVX AI MCP server suitable for clients or workflows that can send HTTP-based MCP requests.
Tools and capabilities
Use find_image_tool when the user describes an outcome rather than knowing a specific product page. Use list_image_tools to inspect all entries in one supported category. Use get_pricing when cost or credit requirements affect the recommendation. All three tools are informational: they provide catalog data and links, not image transformations.
The service is useful for agents that need to recommend a free or appropriately priced browser tool, locate a tool for a particular format, or provide a localized link. It is not a fit when an MCP connection must receive image bytes, perform edits inside the server, or return a transformed file directly.