Kofferly blog post with EU Parliament news article

Kofferly.com

An experiment: Can AI build a complete affiliate content site? Automated research, writing, SEO, publishing. Let's find out.

Visit Live Site 2024 – Present
Load Time
<1s

PageSpeed optimized, Core Web Vitals passed

Google Ranking
3d

First indexed results within 3 days of launch

$ ./the-challenge

The Content Scaling Problem

Affiliate marketing lives on content. But writing good content is time-consuming. SEO built in requires research. Every article needs to hit keywords, be readable, and integrate products naturally. Doing this manually doesn't scale.

"If I'm doing the same thing more than twice, a machine should be doing it."
Experiment
AI

Testing AI content at scale with Claude + Rails

Build Cost
~€5k

What I would charge for a project like this

$ ls ./gallery
$ ./tech-stack
Rails 8 Claude AI MCP DataForSEO Sidekiq MySQL Amazon API
$ ./project-overview

Kofferly is an SEO-optimized affiliate platform with an AI-powered Content Machine. The system researches keywords via DataForSEO, generates bilingual blog posts with Claude AI, scores them across 5 dimensions, and publishes automatically based on priority. From keyword research to finished blog post, everything runs in one pipeline.

Solo Developer 2024 – Present
$ ./content-machine
Feature

Content Machine

The central pipeline: DataForSEO delivers keywords via MCP, Claude generates articles in DE and EN, scores them across 5 dimensions (Featured, Urgency, Freshness, SEO, Brand Fit), and schedules automatic publishing.

  • • Bilingual from single research (DE/EN)
  • • 5-dimension scoring for prioritization
$ ./seo-strategy

SEO Strategy

Every article is optimized for search engines. Keyword research, internal linking, meta descriptions, structured data. All automated, but with manual quality control.

5 Scoring Dimensions
MCP Integration
$ ./affiliate
Feature

Affiliate Integration

Amazon products are embedded via [product:ASIN] placeholders. The system automatically replaces them with product cards including price, ratings, and affiliate links.

  • • [product:ASIN] placeholder system
  • • 100+ products catalogued
$ ./results

Experiment Status

Running

Live experiment with real traffic and SEO data

Ongoing

Continuously improving based on real-world results

What I Learned

Prompt engineering for consistent brand voice is tricky. It takes many iterations until the text sounds natural.

DataForSEO via MCP is powerful. Having keyword research directly in Claude saves enormous time.

Quality control remains important. AI-generated content always needs a human review.

# Similar Projects

RLTracker

Rocket League trading platform

View Project

Splex.gg

NFT rental automation platform

View Project

Golem Overlord

20,000 daily players at peak. Built solo on Hive. Two years of 18-hour days. Honestly the most intense thing I've ever shipped.

View Project
Decorative background

Interested in content automation?

I help build AI-powered content pipelines for SEO and affiliate marketing.

Let's talk