ai-forever/mcp_kandinsky

MCP Kandinsky is a Model Context Protocol server designed to facilitate the generation of images through the Kandinsky neural network, currently supporting version 3.1 via its official API. This server allows users to create images based on textual descriptions, offering a variety of styles to choose from, including DEFAULT, KANDINSKY, UHD, and ANIME. It also supports the use of negative prompts to refine image generation. Users can customize the dimensions of the generated images and save the results to a specified project directory, with options to control file overwriting. To utilize MCP Kandinsky, users must obtain authorization credentials from the FUSION BRAIN website, which are stored in environment variables. The server can be integrated into various agents, such as Cursor or Windsurf, by configuring the agent's settings to include the MCP server details.

"mcpServers": {
    "mcp-kandinsky": {
        "command": "uvx",
        "args": [
            "--from", "mcp_kandinsky", "mcp-kandinsky"
        ],
        "enabled": true,
        "env": {
            "KANDINSKY_API_KEY": "ВАШ_API_KEY",
            "KANDINSKY_SECRET_KEY": "ВАШ_SECRET_KEY"
        }
    }
}