claude-blog-engine
End-to-end blog generation for Claude Code: onboard your business, discover keyword opportunities, and write full SEO articles with images and publishing checklists.
Install
# 1. Clone the repo
git clone https://github.com/maun11/claude-blog-engine.git ~/claude-blog-engine
# 2. Run the install script
bash ~/claude-blog-engine/install.sh
# 3. Restart Claude Code (or start a new session)
That's it. The install script copies 4 skills to ~/.claude/skills/ so they're available in any project.
Skills
| Skill | What it does |
|---|
/blog-hub | Dashboard — shows your status and what to do next |
/blog-onboard [url] | Scrapes your website, builds business profile, finds 3 competitors |
/blog-topics [market] | Keyword research, clustering, scoring, and 10 topic picks |
/blog-write [topic|number] | Full SEO article with images, schema, and publishing checklist |
Quick Start
/blog-hub ← see where you are
/blog-onboard https://yoursite.com ← one-time setup
/blog-topics ← find 10 topics (defaults to US)
/blog-write ← write the top-scored article
/blog-write 3 ← write topic #3 from your pipeline
/blog-write "marketing attribution" ← write a specific topic by keyword
Example Output
Here's a real article written with this engine:
Marketing Attribution Models: What D2C Brands Need to Know
"A practical guide to attribution models for D2C performance marketers. Learn which models work, which lie, and how to get numbers you trust."
Generated from keyword marketing attribution models — includes title, meta description, structured sections, images, and a publish-kit checklist with schema markup.
API Keys
When you run /blog-onboard for the first time, it will automatically create a .env file in your project root with placeholders. Open it, fill in your keys, and reply to continue.
# Required
[email protected] # dataforseo.com — free trial
DATAFORSEO_PASSWORD=yourpassword
ANTHROPIC_API_KEY=sk-ant-... # console.anthropic.com
# Optional — improve article quality
FIRECRAWL_API_KEY=fc-... # firecrawl.dev — better page scraping
TAVILY_API_KEY=tvly-... # tavily.com — deeper topic research
YOUTUBE_API_KEY=AIza... # console.cloud.google.com — video insights
OPENAI_API_KEY=sk-... # platform.openai.com — DALL-E article images
Only DATAFORSEO + ANTHROPIC are required. Optional keys enhance /blog-write — it works fine without them.
You can also copy .env.example from this repo as a starting point.
See docs/apis.md for detailed setup and costs.
How It Works
Three engines run sequentially. Each one feeds the next.
Engine 1 — Onboarding (/blog-onboard)
Runs once. Builds your business profile and finds competitors.
Firecrawl / WebFetch Claude Haiku DataForSEO SERP
Scrape website → Extract business profile → Find competitors
/about, /pricing, product type, ICP, search "{product} software"
/features, /product differentiator, voice filter to 3 direct competitors
Outputs: .claude/blog-config.json with business profile + 3 competitor domains.
Engine 2 — Site Intelligence (/blog-topics)
Runs weekly. Builds your full keyword universe and picks 10 topics.
DataForSEO Claude Haiku DataForSEO
Your rankings (500) + Generate 30 seeds → Expand each seed (×30 parallel)
Competitor keywords based on ICP + pain 30 keywords per seed = ~900
(3 × 200 = 600) points + integrations
DataForSEO bulk KD → Merge + dedup (~2000) → Filter by volume / KD / ranking
~300 actionable keywords remain
Claude Haiku (parallel batches) Claude Sonnet Scoring formula
Classify funnel stage → Group into → opportunity_score =
TOFU / MOFU / BOFU 6–10 clusters log(volume) × difficulty × funnel
50 keywords per call one pillar per cluster 0–100 scale, 70+ = good
Claude Haiku
Select 10 topics → Generate SEO titles → Save to pipeline
one per cluster one Haiku call for all
Outputs: CONTENT-PLAN.md, .claude/blog-keywords.json, .claude/blog-clusters.json, exports/*.csv
Engine 3 — Content Engine (/blog-write)
Runs per article. Research → outline → full article → images → schema.
Step 1 Context Assembly
Pull keyword record + cluster + business profile into one master context object
