home / mcp / seedream mcp server
豆包seedream4.0 MCP, 直接codex, claude code 生图
Configuration
View docs{
"mcpServers": {
"mingforpc-seedream_mcp": {
"command": "uv",
"args": [
"run",
"python",
"-m",
"mcp_doubao.server"
],
"env": {
"ARK_API_KEY": "your-api-key-here"
}
}
}
}You can run a Seedream 4.0 MCP server locally and expose image-generation capabilities via the MCP protocol. It supports automatic image downloads to a chosen directory, configurable output sizes and watermarks, and seamless integration with Claude Code for convenient prompts and batch generation.
You interact with the Seedream MCP server through an MCP client. Start the local server, then request image generation by sending prompts. The server will generate images according to your parameters, save them to your specified folder, and return references or paths to the produced files.
Prerequisites: Python 3.11+, a runtime like uvx/uv, and Claude Code for MCP integration.
1) Clone the project and navigate into the MCP directory.
2) Install dependencies using uvx or pip.
3) Set your API key as an environment variable ARK_API_KEY or configure it in Claude Code.
4) Run the MCP server using one of the supported methods shown below.
Environment variable to set: ARK_API_KEY. Provide your API key from the Volcengine Ark console.
Output and options you can control include image size, watermark, and the number of images generated in a batch. Images are saved locally with auto-generated, non-overlapping filenames.
Example prompts are used to guide Seedream 4.0 in generating visuals. You can request multiple images with varying styles in a single batch.
Common issues include missing dependencies, incorrect ARK_API_KEY, or insufficient write permissions for the output directory. Ensure dependencies are installed, the API key is correct, and the output directory is writable.
If you encounter image overwrite, the server automatically generates unique filenames to prevent collisions.
Asynchronously generate images from a prompt, with options for output directory and image count using the Seedream 4.0 MCP server.