Teardown No. 34 · Social

How Jen Soltys uses n8n and Claude to pull Pinterest quote pins out of her published posts

Jen Soltys wanted several Pinterest pins for every blog post. She says the hard part was never making the images. It was spotting the quotable line in her own writing. She built two n8n workflows with herself in the gap between them. The first sends a published post to the Claude API, which proposes five short quotable lines. Each line lands as a row in Airtable. She ticks the ones she likes, and the second workflow turns only those rows into quote images. She posts the finished pins through Blotato as a separate manual step. She reports 130 candidate lines written since May and 22 turned into images so far.

Jen Soltys · @Jen8nflowsVerified Oct 18, 2026Markdown copy

The problem

She wants several images per post, because “one pin is one shot at getting noticed and five gives a post five chances.” She also says quote pins tend to do well on Pinterest. “The problem was never making the images. It was deciding what to put on them.” When she reread a finished post, every line looked the same to her. She would pick something from the first paragraph or make one pin with the title on it. In her words, the job she handed off “wasn’t the writing. It was the noticing.”

The blueprint (as she describes it)

  1. Keep posts in a table with a checkbox. Her posts live in Airtable with an Extract Quotes checkbox. A second table holds the candidate lines.
  2. Run the extractor. This n8n workflow checks every 30 minutes for posts with Extract Quotes ticked. It fetches the live post and sends the text to Claude asking for five short quotable lines. It writes each line back as its own row. Then it unticks the box on the post so it cannot fire twice on the same one.
  3. Review in the table. The lines sit in rows and nothing happens to them until she looks. She ticks Approve on the ones she likes. She says the checkbox is “the trigger, the queue, and the record of what’s been handled,” and she did not build a screen for it.
  4. Run the generator. This second workflow also checks every 30 minutes, for rows with Approve ticked. It marks each row Generating before doing any work, so a slow run cannot pick up the same row twice. Then it makes the images and marks the row Done. The source does not name the image tool. She says to use “whatever you already use.”
  5. Gate the expensive half. Extraction costs “a few cents” per run. Image generation “costs real money per picture” and only runs on rows she has ticked.
  6. Poll slowly. She first ran both workflows every few minutes. Checking every 5 minutes means “roughly 288 calls a day” to Airtable. She moved to 30 minutes and says nothing about how it feels to use changed. One trigger node is still named “Every 5 Minutes” while running every 30. She advises renaming the node when you change the number.
  7. Post separately. The workflow stops at a finished image attached to a row. She posts through Blotato as a separate step “so a bad run can’t reach anybody.”

Each workflow has twelve nodes. She notes this is separate from another system she uses that writes platform posts and captions.

What she’d do differently: ask for three lines instead of five, start at 30 minutes, set the status guard before the work, and keep the review in a tool she already has open.

Stated results

  • 130 candidate lines written since May. 22 turned into images so far. She says the rest are waiting, not rejected.
  • Of each five lines, “usually a couple are worth using.”
  • Pinterest impressions, clicks or traffic: None stated.
  • Image generation cost: None stated beyond “real money, per image.”

Caveats

  • The post is on her own site. She states at the top that it contains affiliate links. It does not say which links are affiliate links.
  • The page promotes a book she made (“Haul to Sold,” sold on Amazon). It also promotes an affiliate program on tools she builds, paying 20% for up to a year.
  • The page includes a “Build your own” tool. It writes build instructions as a Claude Code prompt or a SKILL.md skill file. The prompt version expects the n8n MCP server to be connected.
  • “Since May” has no year in the source.
  • She says the main value is the noticing, “more than the time saved.” She gives no time-saved figure.
  • She says it is a poor fit if you only publish now and then.
  • Handles: X @Jen8nflows, Instagram @jensoltys, Pinterest pinterest.com/jentriedit (all verified 2026-10-18: linked from jen8n.com/about). LinkedIn, Threads and Bluesky: not checked.

Sources

Want agents like these working your distribution? See the Agent Stack Teardown. Get new teardowns weekly: subscribe (double opt-in).

Get one teardown a week.

Same format, a new operator every issue. Double opt-in.