Volcengine Video Generation MCP Server

MCP (Model Context Protocol) server for Volcengine (Doubao/豆包) Video Generation API - Seedance models.

Features

• ✅ Text-to-Video Generation - Generate videos from text prompts
• ✅ Image-to-Video Generation - Animate images with multiple modes:
  • First frame mode
  • First & last frame mode
  • Reference images mode (1-4 images)
• ✅ Task Management - Query, list, and cancel video generation tasks
• ✅ Comprehensive Parameters - Full control over resolution, ratio, duration, FPS, etc.
• ✅ MCP Resources - Server status, models list, and API documentation

Installation

Prerequisites

• Python 3.11+
• uv (recommended) or pip

Setup

1. Clone the repository:

git clone <repository-url>
cd volcengine-video-mcp

2. Install dependencies:

# Using uv (recommended)
uv sync

# Or using pip
pip install -e .

3. Configure environment variables:

cp .env.example .env
# Edit .env and add your Volcengine API key

Required environment variables:

• ARK_API_KEY: Your Volcengine API key (Get one here)
• ARK_BASE_URL: (Optional) API base URL, defaults to https://ark.cn-beijing.volces.com

Usage

Running the Server

# Using uv
uv run python app/main.py

# Or if installed with pip
python app/main.py

Using with Claude Desktop

Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "volcengine-video": {
      "command": "uv",
      "args": [
        "--directory",
        "/path/to/volcengine-video-mcp",
        "run",
        "python",
        "app/main.py"
      ],
      "env": {
        "ARK_API_KEY": "your_api_key_here"
      }
    }
  }
}

Tools

create_video_task

Create a video generation task.

Example - Text-to-Video:

create_video_task(
    model="doubao-seedance-1-0-pro",
    prompt="一只可爱的猫在草地上玩耍,阳光明媚",
    duration=5,
    ratio="16:9",
    resolution="1080p"
)

Example - Image-to-Video (First Frame):

create_video_task(
    model="doubao-seedance-1-0-pro",
    prompt="镜头慢慢推进,展示更多细节",
    image_url="https://example.com/image.jpg",
    duration=5
)

Example - Image-to-Video (First & Last Frame):

create_video_task(
    model="doubao-seedance-1-0-pro",
    prompt="从白天到夜晚的过渡",
    first_frame_url="https://example.com/day.jpg",
    last_frame_url="https://example.com/night.jpg",
    duration=8
)

Parameters:

• model (required): Model ID (see supported models below)
• prompt (required): Text description for video generation
• image_url: Single image URL for first-frame mode
• first_frame_url: First frame image URL
• last_frame_url: Last frame image URL
• reference_image_urls: List of 1-4 reference images
• resolution: 480p/720p/1080p
• ratio: 16:9/4:3/1:1/3:4/9:16/21:9/keep_ratio/adaptive
• duration: Video duration in seconds (2-12)
• frames: Number of frames (29-289, must satisfy 25+4n)
• fps: Frame rate (16/24)
• seed: Random seed (-1 to 2^32-1)
• camera_fixed: Whether to fix camera
• watermark: Whether to include watermark
• return_last_frame: Whether to return last frame image
• callback_url: Callback URL for task completion notification

get_video_task

Query status and result of a video generation task.

get_video_task(task_id="cgt-2025******-****")

list_video_tasks

List video generation tasks with optional filters.

list_video_tasks(
    page_num=1,
    page_size=10,
    status="succeeded"  # queued/running/cancelled/succeeded/failed
)

cancel_video_task

Cancel or delete a video generation task.

cancel_video_task(task_id="cgt-2025******-****")

Resources

status://server

Get server status and capabilities.

models://list

Get list of supported models and their capabilities.

docs://api

Get comprehensive API documentation.

Supported Models

Seedance Pro Series

• doubao-seedance-1-0-pro - High quality, supports all generation modes, max 12s
• doubao-seedance-1-0-pro-fast - Faster generation, text-to-video and first-frame mode

Seedance Lite Series

• doubao-seedance-1-0-lite-t2v - Text-to-video, max 10s
• doubao-seedance-1-0-lite-i2v - Image-to-video with reference images support

Important Notes

⚠️ Video URL Expiration: Generated video URLs are valid for 24 hours only. Download promptly!

⚠️ Historical Data: Only last 7 days of task history can be queried.

⚠️ Cancellation Limits: Only queued tasks can be cancelled. Running tasks cannot be cancelled.

⚠️ Async Processing: Video generation is asynchronous. Poll task status for results.

Development

Running Tests

# Run all tests
uv run pytest tests/ -v

# Run with coverage
uv run pytest tests/ --cov=app --cov-report=html

Project Structure

volcengine-video-mcp/
├── app/
│   ├── __init__.py
│   ├── main.py              # Main entry point
│   ├── config.py            # Configuration management
│   ├── client.py            # HTTP client for Volcengine API
│   ├── tools/               # MCP tools
│   │   ├── create_video.py
│   │   ├── get_video.py
│   │   ├── list_videos.py
│   │   └── cancel_video.py
│   └── resources/           # MCP resources
│       └── server_info.py
├── tests/                   # Test suite
│   ├── test_config.py
│   └── test_resources.py
├── pyproject.toml          # Project configuration
├── .env.example            # Environment variables template
└── README.md               # This file

Task Status Flow

queued → running → succeeded
                 ↘ failed

queued → cancelled (auto-deleted after 24h)

License

MIT License

Links

• Volcengine Video Generation API Documentation
• FastMCP Documentation
• MCP Specification

Contributing

Contributions are welcome! Please feel free to submit issues and pull requests.

Support

For issues and questions:

• Create an issue in this repository
• Contact Volcengine support for API-related questions