Jimeng

You run a dedicated MCP server that directly delegates image and video generation requests to Jimeng’s AI services. This setup supports standard stdio MCP communication, allows image and video generation with flexible model and parameter options, and can blend local or remote reference images into outputs. You can integrate it with MCP clients like Claude Desktop to generate media on demand.

How to use

Set up the MCP server and connect it to your MCP client (for example Claude Desktop). Once running, you can request image or video generation by calling the server’s exposed tools. You can provide a text prompt, optional reference image (local path or URL), and model/size/quality settings. If you supply a reference image, the server automatically switches to the image blending mode. You can also generate videos with optional first/last frame references.

Typical workflow at a high level:

1. Start the MCP server using the provided stdio configuration. 2) In your MCP client, call the image or video generation tool with your desired parameters. 3) Read back the returned URLs for the generated media and display them in your client.

How to install

Prerequisites: you need Node.js and a package manager (npm or yarn) installed on your system.

Option A: Install via Smithery (recommended for automated client setup)

npx -y @smithery/cli install @c-rick/jimeng-mcp --client claude

Option B: Manual installation

# Install dependencies
yarn install
# or
npm install

Additional configuration and usage notes

Environment variable you’ll need for client configuration: JIMENG_API_TOKEN. This token authenticates requests to Jimeng’s API.

Configure MCP client with the following servers. You can place these configurations in your Claude Desktop or other MCP client settings. A sample Claude Desktop config is shown here for reference.

{
  "mcpServers": {
    "jimeng-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@smithery/cli@latest",
        "run",
        "@c-rick/jimeng-mcp",
        "--key",
        "[Smithery生成]",
        "--profile",
        "[Smithery生成]"
      ]
    }
  }
}

Image generation features

The server exposes a generateImage tool that accepts a prompt and optional parameters to control output quality and size.

Key parameters include:

{
  "prompt": "A realistic sunset over mountains",
  "model": "jimeng-3.0",
  "width": 1024,
  "height": 1024,
  "sample_strength": 0.7,
  "negative_prompt": "blurred, distorted, low quality"
}

Video generation features

The server can generate videos with optional first/last frame references and model selection.

{
  "prompt": "City night timelapse",
  "model": "jimeng-video-3.0-pro",
  "resolution": "1080p",
  "width": 1024,
  "height": 1024,
  "filePath": ["./first.png", "./last.png"]
}

Available tools

generateImage

Generate images from prompts with optional model, size, sampling strength, and negative prompts. Supports image blending when a filePath is provided.

generateVideo

Generate videos from prompts with optional first/last frame references, model, and resolution. Supports customizing output via width/height and filePath for reference frames.