Watch the video with chapters & transcript →
AI TOOLS
In short: Claude generated a complete, importable n8n workflow JSON file from one detailed prompt, no manual node-building required.
Building an n8n workflow by hand means wiring nodes one by one; asking Claude for the JSON file skips straight to a working automation. The stack in this build by Ani Björkström, a Stockholm-based tech consultant, was entirely free: Claude's free tier, a self-hosted local n8n instance, Apify's monthly free credits, and Google Sheets for storage.
Prompt Claude to "create an n8n workflow JSON file for a YouTube scraper that sends data to Google Sheets", followed by a detailed description of what it must do. A detailed prompt raises the odds the JSON works from the beginning; with a short prompt, you upload the file and return to Claude if anything breaks. When Claude finishes, one click on the download button gives you the workflow file ready for import.
Apify is a platform full of ready-made scrapers, and every account receives $5 in free credits each month. The YouTube video and comment scraper costs about $1 per 1,000 results, so the free credits cover roughly 5,000 results every single month. In Apify, open the API button, choose API endpoints, scroll to "run actor synchronously and get key value", and copy that full URL including your token, then tell Claude to call that endpoint in the JSON it writes; never share the token, since anyone holding it can spend your credits.
| Component | Role in the workflow | Cost |
|---|---|---|
| Claude | Generates the importable n8n workflow JSON | Free tier, limited prompts |
| n8n | Runs the automation, self-hosted locally | Free |
| Apify YouTube scraper | Scrapes video data, $1 per 1,000 results | $5 free credit monthly, about 5,000 results |
| Google Sheets | Stores results, deduplicated on the ID column | Free |
In n8n, click the plus button for a new workflow, open the three-dots menu, choose "import from file", and select the JSON Claude produced. Executing the workflow opens a form with the scraper's parameters: search term (n8n in the demo), max videos per search, shorts and streams set to zero to exclude them, results sorted by view count, and any upload date, content type, and duration. The demo run scraped four n8n videos from YouTube and saved them straight into the connected sheet.
Set the Google Sheets node's "column to match on" to ID, so every incoming record is checked against existing rows before insertion. If the ID already exists the row is skipped; if not, it is added, which keeps repeated runs from filling the sheet with duplicates. The sheet's columns, ID, URL, title, channel name, and so on, mirror the fields n8n receives, and any unmapped field can be fixed by dragging and dropping it into place.
Yes. Claude's free tier generates the JSON, n8n runs self-hosted on a local computer at no cost, and Apify's $5 monthly free credit covers about 5,000 YouTube results.
Go back to Claude, describe the error, and ask it to adjust the JSON. Then re-import the corrected file into n8n.
The "run actor synchronously and get key value store" endpoint, copied with the token included, pasted into the workflow's URL field.
Read the full video transcript, with chapters, on the video page →
Want this working inside your finance team?