Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@volcengine-seedream-img-mcp Generate a realistic portrait of a young woman with flowing hair."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for Volcengine Ark Doubao-Seedream image generation and editing.
Tool | Description |
| Text-to-image or image-to-image (1–14 refs). Returns |
| Edit/restyle a single image with a prompt (seededit-3.0). Returns |
| Generate a group of related images. Returns |
| Poll a previous job by |
Related MCP server: Seedream 4.0 MCP Server
Add to .mcp.json:
{
"mcpServers": {
"seedream": {
"command": "npx",
"args": ["-y", "volcengine-seedream-img-mcp"],
"env": { "ARK_API_KEY": "your_key_here" }
}
}
}Set ARK_API_KEY environment variable, or pass --api-key <value> as a CLI flag.
Get your key at: https://console.volcengine.com/ark/region:ark+cn-beijing/apiKey
Model ID | Resolution support |
| 1K, 2K, 4K |
| 2K, 4K |
| 2K, 3K |
1Kresolution is silently upgraded to2Kwhen using a model that doesn't support it.
edit_image always uses doubao-seededit-3-0-i2i (model not selectable).
generate_image / edit_image / generate_image_group
→ { request_id }
→ poll get_image_result until state = "succeeded"
→ images[].url (download within 24h)MIT
Edit or restyle a single reference image using a text prompt (uses doubao-seededit-3-0-i2i). Output size is adaptive to input image aspect ratio. Returns a request_id to poll with get_image_result.
Inputs: image (required), prompt (required), watermark, response_format.
Generate a single image from a text prompt (text-to-image), or using 1–14 reference images (image-to-image). Returns a request_id to poll with get_image_result.
Inputs: prompt (required), image, model, watermark, +3 more.
Generate a group of related images from a prompt and optional reference images. Returns a request_id to poll with get_image_result.
Inputs: prompt (required), image, model, watermark, +4 more.
Poll the result of a previous generate_image, edit_image, or generate_image_group call. Returns state: pending | succeeded | error.
Inputs: request_id (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 4 tools
Each tool targets a distinct operation: editing a single image, generating a single image, generating a group, and polling results. No overlap in purpose.
All tools follow a consistent verb_noun pattern in snake_case: edit_image, generate_image, generate_image_group, get_image_result.
4 tools is well-scoped for an image generation server, covering core creation, editing, and result polling without unnecessary extra tools.
The set covers the main workflow (generate, edit, poll), but is missing optional features like cancellation or listing pending requests, which are minor gaps.
LLM chat, text tools, image generation, editing, batch image jobs, and asynchronous video generation
Create images and videos from prompts, with options for image mixing, reference images, and start/…
Generate and edit images, create videos, quote credit costs, and retrieve private results.
Generate, edit and upscale images with references, batch jobs, cost previews and original downloads.