douyin-mcp-server

The system automatically handles large audio files that exceed API limits (typically 1 hour or 50MB) using the following workflow:

1. Detects audio duration and file size.
2. Uses FFmpeg to split the audio into 9-minute segments.
3. Calls the API to transcribe each segment sequentially.
4. Merges all transcribed text results into a single output.

The transcript extraction feature requires a SiliconFlow API key. You must set this as an environment variable named API_KEY before running the extraction commands.

1. Obtain a key from https://cloud.siliconflow.cn/
2. Set the environment variable:

export API_KEY="your-siliconflow-api-key"

To use this project as an MCP server in Claude Desktop or Cherry Studio, add the following configuration to your MCP settings file.

Note: You can use a SiliconFlow key via API_KEY or an Alibaba Cloud DashScope key via DASHSCOPE_API_KEY.

{
  "mcpServers": {
    "douyin-mcp": {
      "command": "uvx",
      "args": ["douyin-mcp-server"],
      "env": {
        "API_KEY": "sk-xxxxxxxxxxxxxxxx"
      }
    }
  }
}

The WebUI is the recommended way for general users to interact with the service via a browser.

Prerequisites

• uv (Python package manager)
• Python 3.10+
• FFmpeg (for audio/video processing)

Setup Steps

1. Clone the repository:

   git clone https://github.com/yzfly/douyin-mcp-server.git
   cd douyin-mcp-server

2. Install dependencies using uv:

   uv sync

3. Start the service:

   uv run python web/app.py

4. Access the interface at http://localhost:8080.

# 1. 克隆项目
git clone https://github.com/yzfly/douyin-mcp-server.git
cd douyin-mcp-server

# 2. 安装依赖
uv sync

# 3. 启动服务
uv run python web/app.py

To use the douyin-video tool, you need to install the required Python packages and ensure FFmpeg is installed on your system for audio/video processing.

Python Dependencies

pip install requests ffmpeg-python

System Requirements (FFmpeg)

• macOS: brew install ffmpeg
• Ubuntu: apt install ffmpeg

To extract text (AI voice recognition), you must provide an API Key from SiliconFlow. There are two ways to configure it:

Option 1: Via WebUI (Recommended)

1. Open the WebUI in your browser.
2. Click the 「API 未配置」 (API not configured) button at the top.
3. Enter your API Key and save. The key is stored locally in your browser.

Option 2: Via Environment Variable

Set the API_KEY environment variable before running the server:

export API_KEY="sk-xxxxxxxxxxxxxxxx"
uv run python web/app.py

To run the WebUI, set your API_KEY environment variable and execute the app.py script. The service defaults to http://localhost:8080 unless HOST or PORT environment variables are provided.

export API_KEY="sk-xxx"
python web/app.py

To use this server with Claude Desktop, add the following configuration to your claude_desktop_config.json. Ensure you replace your-siliconflow-api-key with your actual key.

{
  "mcpServers": {
    "douyin-mcp": {
      "command": "uvx",
      "args": ["douyin-mcp-server"],
      "env": {
        "API_KEY": "your-siliconflow-api-key"
      }
    }
  }
}

The Douyin MCP Server supports two Automatic Speech Recognition (ASR) backends. You must set one of the following environment variables to enable text extraction features:

1. SiliconFlow (Recommended):

   • Set API_KEY with your key from https://cloud.siliconflow.cn.
   • Default model: FunAudioLLM/SenseVoiceSmall.

2. Alibaba DashScope (Aliyun Bailian):

   • Set DASHSCOPE_API_KEY with your key from Alibaba Cloud.
   • Default model: qwen3-asr-flash.

Note: get_douyin_download_link and parse_douyin_video_info do not require an API key.

Link Parsing Failed

• Ensure the link is a valid Douyin share link (e.g., https://v.douyin.com/xxxxx/ or a full video URL).

Transcript Extraction Failed

• Verify that the API_KEY environment variable is set.
• Check if the API key is valid and has sufficient quota.
• Ensure FFmpeg is installed and accessible in your system path.

Slow Download Speeds

• This depends on your network conditions and video size; the script will display download progress.

When integrated via MCP, the following tools are available: