Hey everyone! 
If you build AI agents or RAG pipelines in n8n, you’ve probably faced this: fetching raw HTML with standard HTTP nodes fills your LLM context window with navigation menus, footer links, and cookie banners—burning up to 80% of your tokens unnecessarily.
To solve this for my own automation workflows, I built a lightweight API called WebPulse (hosted on Vercel and available on RapidAPI), and I wanted to share a simple, reusable n8n pattern that you can paste directly into your canvas.
What this workflow does:
- Takes any input URL.
- Sends it through WebPulse’s
/api/v1/markdownendpoint. - Returns:
- Clean GitHub-flavored Markdown.
- Formatted plain text.
- Pre-split semantic
rag_chunksready to pipe directly into Vector Store or OpenAI/Claude nodes. - Exact token savings percentage (
llm_token_savings_percent).
Copy-Paste n8n Workflow JSON:
You can copy this JSON block and paste it directly into your n8n editor canvas (Ctrl + V):
How to test it for free:
-
Free Tier on RapidAPI (50 requests/month included for testing):
https://rapidapi.com/Juanamigazo/api/webpulse-ai-markdown-seo-audit-and-contacts -
Live Browser Playground (no signup needed to test a URL):
https://webpulseapi.vercel.app
I’m an indie builder experimenting and building cloud tools in public from Santa Fe, Argentina. I would love to hear feedback from the n8n community—what additional data fields or extraction features would help your daily workflows?