6 min read
n8n + Claude API: automating competitive monitoring
How I build an n8n workflow that scrapes, summarizes and synthesizes daily competitive monitoring with the Claude API.
n8nclaudeautomationai
The goal
Get a synthetic summary every morning of competitive announcements: sites, blogs and release notes browsed overnight, with the points worth a closer look.
Workflow architecture
Three layers in n8n:
- Collection: HTTP Request nodes scheduled via Cron
- Synthesis: HTTP node calling the Claude API with a structured prompt
- Distribution: Slack or email depending on preferred channel
The prompt that works
I've iterated quite a bit. Three things that make a difference:
- Give a clear role: "You're a product analyst tracking competitor X..."
- Force a strict output format (JSON or sectioned Markdown). Without it, Claude varies its structure.
- Set an explicit length limit ("3 bullets max per item")
Cost in practice
Over a month of daily execution:
- Haiku: ~USD 0.30/month for ~50k input tokens per day
- Sonnet: ~USD 1.50/month for the same volume
- Opus: ~USD 7/month (reserve for analyses that really warrant it)
For pure monitoring, Haiku is enough. Sonnet when you want a real argued synthesis.
What not to do
Don't ask Claude to scrape. It's slow and unreliable. n8n has dedicated nodes (HTTP Request, HTML Extract) that do it better and cheaper.