Jimeng4

This MCP server enables you to access 即梦AI's multimodal generation capabilities through MCP clients such as Cursor and Claude Desktop. It supports the latest 即梦4.x and related tools, runs across macOS, Linux, Windows, and WSL, and can be used as a standalone library. You can submit image and video generation tasks, monitor asynchronous results, and integrate generation workflows directly into your pipelines.

How to use

You connect to this MCP server from an MCP client by configuring a local or remote MCP entry. The server exposes a set of image and video generation tools you can invoke from the client, with asynchronous task handling where supported. You can submit generation requests, check task progress, and retrieve results when ready. Use the available tools to generate images from text or edits from existing images, and to create videos from textual prompts or image sequences.

How to install

Prerequisites you need before installing: Node.js (version 14 or higher) and npm. You may also install Python or other runtimes if you plan to extend integrations, but the MCP server itself runs in Node.js.

Configuration and usage notes

Environment variables you must set before running the server are the access keys for 即梦AI services.

# macOS/Linux or WSL
export JIMENG_ACCESS_KEY=YOUR_ACCESS_KEY
export JIMENG_SECRET_KEY=YOUR_SECRET_KEY

# Windows PowerShell
$env:JIMENG_ACCESS_KEY = "YOUR_ACCESS_KEY"
$env:JIMENG_SECRET_KEY = "YOUR_SECRET_KEY"

Quick start commands

Install the MCP server globally for quick access across platforms, or clone and build locally for development.

# Global installation
npm install -g jimeng4-mcp

# Or local development setup
git clone <your-repo-url>
cd jimeng4-mcp
npm install
npm run build

MCP client configuration examples

Configure clients like Cursor or Claude Desktop to load the MCP server via a local stdio process. The following examples show how to start the server using the recommended npx invocation and environment variables.

{
  "mcpServers": {
    "jimeng": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "jimeng4-mcp"
      ],
      "env": {
        "JIMENG_ACCESS_KEY": "YOUR_ACCESS_KEY",
        "JIMENG_SECRET_KEY": "YOUR_SECRET_KEY"
      }
    }
  }
}

Available tools overview

The MCP server supports a suite of image and video generation tools. You can invoke these tools from MCP clients and manage asynchronous tasks when supported.

Troubleshooting

If you encounter authentication issues, verify that JIMENG_ACCESS_KEY and JIMENG_SECRET_KEY are correctly set in your environment and in your MCP client configuration.

If you see issues starting the stdio server, ensure the command path is correct (for example, using npx or full Node.js paths) and that the environment variables are visible to the client process.

Docker deployment

You can package the MCP server in a Docker container for consistent runtimes across environments.

FROM node:18-alpine

RUN npm install -g jimeng4-mcp

ENV JIMENG_ACCESS_KEY=YOUR_ACCESS_KEY
ENV JIMENG_SECRET_KEY=YOUR_SECRET_KEY

CMD ["jimeng4-mcp"]

Development and publishing

For local development, run in dev mode to test changes, then build and publish as needed.

# Development flow
npm run dev
npm run build
npm test

# Publish a new npm version
npm version patch|minor|major
npm publish

Available tools

jimeng-v40-generate

4.0 image generation tool that supports text-to-image, image editing, and multi-input generation with up to 10 inputs and 15 outputs, including 4K output.

jimeng-i2i-v30

image-to-image tool that edits images based on text prompts (add/delete entities, change style/color/background, etc.).

jimeng-t2i-v31

text-to-image tool with upgraded visual quality, broader style variations, and richer details.

jimeng-t2i-v30

text-to-image tool focused on improved text responsiveness and support for artistic fonts.

generate-video

text-to-video tool that generates videos from prompts; supports asynchronous mode by default.

submit-video-task

submit an asynchronous video generation task and obtain a task identifier.

get-video-task

retrieve the result of a previously submitted video task.

generate-image

legacy image generation tool that accepts textual prompts and optional illustration and color parameters.