A zero-touch n8n cost scanner reads token usage straight from n8n's execution metadata to track AI API spending (Claude, Perplexity, Firecrawl) across every workflow, without touching existing workflows. It logs costs to Google Sheets, sends a weekly cost digest, and fires budget alerts before a surprise bill lands, part of a production n8n workflows series.
What you will be able to do
- Build a single n8n scanner workflow that reads token usage from execution metadata for every AI workflow, without modifying any of them
- Set up a Google Sheets dashboard that shows monthly totals, daily trends, and top-spending workflows
- Configure budget alerts that fire once a day when spending crosses a threshold, instead of per execution
- Generate a weekly email digest that breaks down cost by workflow and by model
- Estimate hidden costs beyond API tokens, including n8n execution limits, data scraping, storage, and maintenance hours, so you can plan a realistic monthly budget
Before you start
- An n8n instance (self-hosted or Cloud) with existing AI workflows using nodes like OpenAI, Anthropic, or Google Gemini
- Access to n8n's REST API to list and fetch execution data
- A Google Sheets account to receive the cost log and dashboard
- Basic familiarity with n8n workflow building (nodes, triggers, HTTP requests)
Reference
| Cost item | Rate | Real-world translation |
|---|---|---|
| Claude Sonnet 4.6 | $3 per million input tokens, $15 per million output tokens | 1,000 content generation calls is about $30; 10,000 calls is about $300 |
| Perplexity API | $5 per 1,000 searches | 100 queries/week is about $2/month; 500 queries/week is about $10/month |
| n8n Cloud Starter | 2,500 executions/month, $24 | A workflow running every 15 minutes (96/day) is about 2,880 executions/month, over the limit |
| n8n Cloud Pro | 10,000 executions/month, $60 | Covers higher-frequency workflows that outgrow Starter |
| Firecrawl | $19 per 3,000 pages, varies by plan | 500 competitor sites scraped daily is about 150,000 pages/month, needs $99-$399 plans |
| Apify | $29/month Starter (includes $29 in credits), $0.30 per compute unit | Scales with scraping volume and compute-heavy jobs |
| Airtable Free | 1,000 records per base | 100 executions/day logged is about 3,000 records/month, exceeds free tier in about 10 days |
| Airtable Team | 50,000 records per base, $24/user/month | 500 executions/day is about 15,000 records/month |
Common errors and fixes
| What goes wrong | The fix |
|---|---|
| You add tracking nodes manually to each workflow and forget to update new ones | Build one scanner workflow that reads token usage from n8n execution metadata for every workflow automatically |
| A workflow you forgot about, like an hourly competitor scan, quietly costs $75/month | Run the scanner's per-workflow cost report to spot and pause low-value automations |
| A low-frequency workflow that runs every 5 minutes burns through n8n's execution limit even though it delivers little value | Track execution count against value delivered, not just API cost, and adjust trigger frequency |
| The scanner only captures LLM calls, missing Perplexity, Firecrawl, and other external API costs, about 20 percent of total spend | Add the optional manual tracking pattern from Step 7 of the article for non-AI API costs |
| Google Sheets or Airtable logging hits free-tier record limits faster than expected as execution volume grows | Plan for a paid tier, such as Airtable Team at 50,000 records, before volume outgrows the free plan |
Read the full walkthrough
The complete lesson, with screenshots and any downloads, is published on Substack as part of From Demo to Dependable: Production n8n Workflows.
More in this section
Continue the course
Browse all lessons in the From Demo to Dependable: Production n8n Workflows course, or subscribe to the GenAI Unplugged newsletter to get new lessons in your inbox.
