I’m excited to share my n8n Xiaohongshu agent project and the tutorial video I recorded.

What I Built

This assistant automatically detects new photos added to a watched directory and turns them into ready-to-publish Xiaohongshu content.

The workflow is:

1. Watch a local directory for newly added photos
2. Send each image to a multimodal LLM (specifically GPT-4o) for a detailed visual description
3. Pass that description to an agent that writes a polished Xiaohongshu bio/description and relevant tags
4. Publish the post to my Xiaohongshu profile through the Xiaohongshu MCP server

This setup removes the repetitive posting steps and keeps content quality high by combining visual understanding with automated copy generation.

Video Tutorial

Why This Is Useful

• Speeds up photo-to-post publishing
• Improves caption quality with multimodal context
• Keeps hashtags and descriptions consistent
• Demonstrates practical MCP + n8n automation in production-like workflows