Crawl4AI with n8n: web pages as Markdown in your workflows

1. Run n8n and Crawl4AI side by side

services:
  n8n:
    image: n8nio/n8n
    ports: ["5678:5678"]
  crawl4ai:
    image: unclecode/crawl4ai:latest
    shm_size: "1g"

In the same compose network, n8n reaches Crawl4AI at http://crawl4ai:11235.

2. HTTP Request node

• Method: POST
• URL: http://crawl4ai:11235/md
• Body content type: JSON

{
  "url": "{{ $json.url }}",
  "f": "fit"
}

The response contains the Markdown in the field markdown. Use it in the next node as {{ $json.markdown }}.

3. Typical workflows

• Research agent: Webhook → HTTP Request (Crawl4AI) → AI Agent node summarises the page.
• Knowledge base: Sitemap URLs → Loop → Crawl4AI → split into chunks → vector store.
• Monitoring: Schedule → Crawl4AI → compare with last version → Slack or email on change.

There are also unofficial community nodes for Crawl4AI. The HTTP approach above uses only the official API and keeps working when the community node is not updated.