Wanyi-Watermark

You can run a Model Context Protocol (MCP) server that parses Douyin and Xiaohongshu links to extract watermark-free videos and images, and also perform video transcription. This server automatically detects content type, handles multiple platforms, and requires no API keys for link parsing itself. It is designed to be integrated into MCP-enabled clients to simplify obtaining media assets from short-video and note-sharing links.

How to use

You deploy the wanyi-watermark MCP server locally or in a compatible MCP environment and connect to it through your MCP client. Use the provided tools to parse sharing links from Douyin or Xiaohongshu, which will return either a watermark-free video download URL or a set of images for graphic notes. If transcription is needed, you can extract text from video using the built-in transcription tool after configuring the optional API key for transcription.

How to install

Prerequisites: you need Python 3.10 or newer. You should also have access to an MCP-capable client where you can register an MCP server.

Configuration and startup details

Configure your MCP client to connect to the local wanyi-watermark MCP server using a standard stdio integration. The server exposes a local process you run via the MCP framework. If you plan to enable video transcription, obtain an API key from the transcription service and provide it as an environment variable.

Tools and capabilities

The MCP server provides the following tools for end users:

• parse_douyin_link: Accepts a Douyin share link or text containing the link and returns either a watermark-free video JSON or a list of images for notes depending on content type.
• parse_xhs_link: Accepts a Xiaohongshu share link or text containing the link and returns a watermark-free video URL or a set of images in WebP and PNG formats for notes.
• parse_generic_link: Accepts any share link or text containing links and returns a JSON string with resource links and information.
• extract_douyin_text: Extracts spoken content from a Douyin video, with optional model selection. Requires a transcription API key.

Security and credentials

Link parsing itself does not require an API key. If you enable video transcription, you must provide a valid API key for the transcription service (DASHSCOPE_API_KEY). Keep API keys secure and do not expose them in client configurations.

Troubleshooting and notes

If parsing fails on a platform, the server will automatically attempt a generic parsing mechanism as a fallback. Ensure your API key (if used for transcription) is valid and that the MCP client is correctly configured to invoke the provided tools.

Example configuration and runtime guidance

Below is the core runtime configuration to register the MCP server in an MCP-enabled client that supports stdio-based servers. The command uses uvx to start the wanyi-watermark server and passes the required environment variable for transcription when needed.

{
  "mcpServers": {
    "wanyi-watermark": {
      "command": "uvx",
      "args": ["wanyi-watermark"],
      "env": {
        "DASHSCOPE_API_KEY": "sk-xxxx"
      }
    }
  }
}

Available tools

parse_douyin_link

Parses a Douyin share link or text containing the link and returns either a watermark-free video download URL or a set of images for notes depending on content type.

parse_xhs_link

Parses a Xiaohongshu share link or text containing the link and returns a watermark-free video URL or a set of images in WebP and PNG formats.

parse_generic_link

Parses any share link or text containing links and returns a JSON string with resource links and information.

extract_douyin_text

Extracts spoken content from a Douyin video, with optional transcription model selection. Requires a transcription API key.