I built a tool that writes cold email openers from any URL - here's what I learned

Published: (February 19, 2026 at 03:24 PM EST)
2 min read
Source: Dev.to

Source: Dev.to

The Problem with Cold Outreach

Cold outreach is a grind. The part that always slowed me down was writing that first line in an email—the one that actually references something specific about the company so it doesn’t look like a template blast.

I spent way too long doing this manually: open their website, read for 2–3 minutes, write something. Multiply that by 50 prospects and you’ve burned half your day before sending anything.

Introducing Icebreakr

I built Icebreakr to automate the creation of personalized opening lines.

  • Input: Paste a company’s website URL (or LinkedIn profile).
  • Process:
    1. Fetch the page and grab the title and meta description.
    2. Pass that context to gpt‑4o‑mini with a prompt tuned for a cold‑email tone.
  • Output: Returns five opening lines you can pick from or riff on.

Simple Tech Stack

  • Next.js 14
  • Tailwind CSS
  • OpenAI API
  • Stripe (for payments)

Example Output

For linear.app the tool generated:

  1. “I noticed Linear’s focus on integrating AI into product development and thought it aligns well with current industry needs.”
  2. “Your approach to streamlining product planning with AI is impressive and highlighted in your recent updates.”

The results aren’t always perfect, but they usually provide a solid starting point that saves time—even when you need to tweak them.

Limitations

  • Meta description matters: When a site has a clear meta description, the outputs are more specific. Sites without one fall back to just the page title, limiting specificity.
  • LinkedIn URLs: LinkedIn blocks scraping, so the tool falls back to domain information when scraping fails.

Pricing & Availability

Call for Feedback

I’d love feedback, especially from people who do outreach regularly.

  • Are the openers specific enough?
  • Too generic?
  • What would actually make you pay for something like this?
0 views
Back to Blog

Related posts

Read more »

Apex B. OpenClaw, Local Embeddings.

Local Embeddings para Private Memory Search Por default, el memory search de OpenClaw envía texto a un embedding API externo típicamente Anthropic u OpenAI par...