Jimeng AI

You can run the multimodal MCP server locally and connect to it from MCP clients to generate images and videos from text prompts. This server supports asynchronous task handling, cross‑platform environments, and thorough parameter control so you can tailor outputs to your needs.

How to use

You will run a local MCP server that exposes tools for image and video generation. In MCP clients, connect to the local stdio server you start or to a remote HTTP endpoint if you have one. Use the available tools to create images from text, generate videos from prompts, submit long video tasks, or fetch results for asynchronous tasks. The server handles authentication keys via environment variables, so set your access keys before starting.

How to install

{
  "mcpServers": {
    "jimeng_ai_mcp": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "jimeng-ai-mcp"],
      "env": {
        "JIMENG_ACCESS_KEY": "YOUR_ACCESS_KEY",
        "JIMENG_SECRET_KEY": "YOUR_SECRET_KEY"
      }
    }
  }
}

Additional configuration and usage notes

This server supports multiple platforms and provides a single, consistent MCP interface for both image and video generation. Ensure you provide the required access keys as environment variables. When configuring clients, you can place these keys in a local environment or a dedicated .env file in your working directory so the MCP client can read them automatically.

Security and environment variables

Protect your access keys. Do not commit them to source control. Use environment variables JIMENG_ACCESS_KEY and JIMENG_SECRET_KEY to authenticate with the generating services. You can set these variables in your shell session, in a per‑project env file, or in your system environment depending on your deployment method.

Examples of common tasks

Generate an image from text using the image generation tool. Generate a short video from a descriptive prompt using the video generation tool. Submit a video task to run asynchronously and check its status later.

Troubleshooting and tips

If you encounter issues starting the MCP server, verify that Node.js and npm are installed and that you can run npm commands. Ensure your environment variables are set in the runtime context where you start the server. For asynchronous video tasks, plan to poll for results using the provided task ID and design your workflow to handle pending and running states.

Notes on deployment

You can run the MCP server in your local development environment or deploy via Docker or your favorite orchestration platform. The configuration shown provides a ready‑to‑use stdio setup with the necessary command and environment variables to start generating content immediately.

Available tools

generate-image

Create high-quality images from text prompts with optional attributes like illustration, color, and aspect ratio.

generate-video

Produce videos from descriptive prompts with support for asynchronous processing and status tracking.

generate-image-to-video

Convert a static image into a dynamic video with optional animation prompts and aspect ratio.

submit-video-task

Submit a long-running video generation task for asynchronous processing.

get-video-task

Query the status and results of an asynchronous video task by its ID.