Content SEO Orchestrator

A Claude Code–native content SEO pipeline. Research → Brief → Write → SEO Score → Publish → IndexNow in 15 min/day at ~$0.45/post. MIT-licensed. Fork it, adapt it, ship it.

What's new in v1.0.0 (2026-05-19): 4 vertical starter pipelines (B2B SaaS, D2C, local biz, agency) · /seo-refresh command · standalone AEO Scorecard · feature comparison vs Profound/Peec/AI Peekaboo · one-shot setup.sh · full CHANGELOG.

Table of contents

• Real-world result
• What this is
• Why this exists
• Commands
• Cost per post
• How this compares to alternatives
• Features at a glance
• Architecture
• Quick start
• Starter pipelines by vertical
• Roadmap
• Docs deep-dives
• FAQ
• Contributing
• License

Real-world result

One brand we run this pipeline for went from 131 → 964 avg impressions/day in 12 months — a 7.3× lift. Monthly impressions went from 2,142 → 39,240 (18× growth). Blog content produced by this orchestrator accounted for 51.8% of all impressions (126,829 of 244,861) across 119 posts.

Honest caveat: clicks didn't grow proportionally over this window — CTR dropped because new content ranked but the title/meta layer hadn't been tuned yet. That gap is now closed by the new /seo-refresh command which re-optimizes live posts using GSC data.

Top posts produced by the pipeline (real, live):

This is the Asva AI blog content (12 posts on AEO / AI visibility) shipped as the working example in /output/ and pipeline.yaml. Browse those to see real input → real output.

What this is

A 7-step content SEO pipeline that runs as Claude Code slash commands, with 4 human gates so you stay in control of the high-judgment calls.

Research → Brief → Write → SEO Score → Cover/Diagrams → Publish → IndexNow → Refresh
   GATE     GATE    GATE                                  GATE                 GATE

You handle: research angle, brief approval, copy voice, publish decision, refresh decision.
The pipeline handles: structure, templates, SEO scoring, image generation, CMS publishing, instant indexing, CTR optimization.

Why this exists

Most SEO/AEO content stacks fail one of two ways:

1. Fully automated — thin, generic posts you wouldn't put your name on. Google's March 2026 core update gutted exactly this kind of content.
2. Fully manual — 80% of your day is research, formatting, and CMS clicking. You don't scale.

This pipeline keeps the boring stuff automated (templates, scoring, CMS payloads, image generation, indexing) and the interesting stuff human (research angle, brief approval, copy voice, publish decision).

Commands

Cost per post

Full breakdown in config/api-setup.md.

How this compares to AI visibility tools

This is a content production pipeline. It's complementary to AI tracking tools, not a replacement.

If you want to know whether you show up in ChatGPT, use a tracker. If you want to make the posts that get you cited, use this. Full feature matrix: docs/COMPARISON.md.

Features at a glance

Research

• Perplexity Deep Research API integration (sonar-deep-research, ~$0.45/post)
• Manual paste fallback if no API key
• Source URLs preserved on every stat extracted
• Cost tracking on every call

Content

• Brand voice + tone driven by config/seo-settings.yaml — no hardcoded brand
• Brief-first writing (no draft until brief is approved)
• Multi-brand: keep one repo, swap settings per project (config/seo-settings.example.yaml ships with the Asva AI reference brand)
• Real examples enforced; hypotheticals banned

SEO + AEO scoring

• 10-check rubric (see docs/AEO-SCORECARD.md)
• Scores for both SERP rank and LLM citation likelihood
• Specific fix suggestions when score < 80

Images

• generate_covers.py — branded 1200×630 covers via Pillow
• generate-blog-images.py — HTML→PNG via headless Chrome (hero + OG + Twitter + LinkedIn cards per post)
• Figma/FigJam infographics via Figma MCP (Mermaid → FigJam)

Publishing (Sanity reference implementation)

• Direct HTTP API, not the Sanity MCP. Why: the MCP's create_documents_from_json overwrites your custom _id with a random UUID, breaking deterministic frontends. Direct curl preserves _id like blogPost-<slug>.
• IndexNow ping for Bing/Yandex/Seznam/Naver after live publish
• Pre-flight checks: SEO score gate, required-field validation, human confirmation

Refresh / CTR optimization

• /seo-refresh reads GSC data, diagnoses high-impressions/low-CTR pages, proposes title/meta/intro changes
• Categorizes the problem (title mismatch vs intent drift vs position-floor) so you know what kind of fix to apply

Pluggable CMS

Sanity ships as the reference. Swap in WordPress / Contentful / Strapi / Webflow / Ghost by editing .claude/commands/seo-publish.md. Pipeline, briefs, writing, scoring, images stay CMS-agnostic.

Architecture

Full deep-dive: docs/ARCHITECTURE.md.

your-project/
├── pipeline.yaml                  # All posts + status (single source of truth)
├── .env                           # API keys (gitignored)
├── config/
│   ├── seo-settings.yaml          # Your brand, CMS, SEO rules
│   ├── seo-settings.example.yaml  # Reference brand (Asva AI)
│   └── api-setup.md
├── templates/                     # Research brief, content brief, blog post, CMS payload
├── examples/                      # Starter pipelines by vertical
│   ├── b2b-saas.pipeline.yaml
│   ├── d2c-ecommerce.pipeline.yaml
│   ├── local-business.pipeline.yaml
│   └── agency.pipeline.yaml
├── docs/                          # Deep-dive references
│   ├── ARCHITECTURE.md
│   ├── COMPARISON.md
│   ├── AEO-SCORECARD.md
│   └── FAQ.md
├── output/<slug>/                 # Per-post working dir
├── covers/                        # 1200x630 branded covers
├── blog-images/<slug>/            # Per-post social images
├── infographics/                  # FigJam diagram exports
├── generate_covers.py             # Pillow cover generator
├── generate-blog-images.py        # HTML→PNG social card generator
├── setup.sh                       # One-shot bootstrap
└── .claude/commands/              # The /seo-* slash commands

Quick start

Fastest path (~3 minutes):

git clone https://github.com/viren040/content-seo-orchestrator.git
cd content-seo-orchestrator
./setup.sh

The bootstrap script creates a venv, installs deps, copies .env.example → .env, and walks you through what to edit next.

Manual path (if you'd rather):

# 1. Clone
git clone https://github.com/viren040/content-seo-orchestrator.git
cd content-seo-orchestrator

# 2. Configure
cp .env.example .env
# Edit .env — at minimum: PERPLEXITY_API_KEY, SANITY_AUTH_TOKEN

# 3. Edit your brand
# config/seo-settings.yaml — brand.name, brand.website, sanity.project_id, voice, content_pillars

# 4. Pick a starter pipeline
cp examples/b2b-saas.pipeline.yaml pipeline.yaml
# (or d2c-ecommerce / local-business / agency)

# 5. (Optional Python deps for image generation)
python -m venv .venv && source .venv/bin/activate
pip install pyyaml pillow requests

# 6. In Claude Code:
/seo-status      # dashboard
/seo-daily       # auto-picks next post, runs next step, pauses at human gates

Starter pipelines by vertical

Four pre-built pipeline.yaml files in examples/, each with 8 starter posts:

Copy the closest fit to ./pipeline.yaml, edit slugs/keywords/internal links, run /seo-status.

Roadmap

Shipped in v1.0.0:

• Perplexity Deep Research API
• Direct Sanity HTTP publishing (with the MCP _id workaround)
• IndexNow ping
• Cover + social image generators
• Multi-brand config
• /seo-refresh for CTR re-optimization
• 4 starter pipelines by vertical
• AEO Scorecard (10-check rubric)

Confirmed next:

• DataForSEO MCP integration — automated keyword opportunity detection
• GSC API integration — automatic data fetch for /seo-refresh
• Cannibalization audit script
• Multi-language content (one source draft → localized variants)

Likely:

• WordPress / Contentful / Webflow publishers (PRs welcome)
• Rank tracking dashboard
• AI citation tracking (which LLMs cite the post)

Want one of these sooner? Open a Discussion or send a PR. See CONTRIBUTING.md.

Docs deep-dives

FAQ

A few quick answers; full FAQ in docs/FAQ.md.

Is it really $0.45/post? Yes — that's the marginal API cost per post (one Perplexity Deep Research call). You also need a Claude Code subscription, which you presumably have if you're reading this. CMS publishing is free tier. IndexNow is free.

Do I need to use Claude Code? Yes — the slash commands run inside Claude Code. Porting to another LLM client would be straightforward (each command is just a markdown prompt), but we don't ship one.

Do I need Sanity? No — Sanity is the reference CMS. Edit one file (.claude/commands/seo-publish.md) to swap to WordPress / Contentful / Strapi / Webflow / Ghost.

Can I run this for multiple clients? Yes — fork the repo per client. Each gets its own config/seo-settings.yaml + pipeline.yaml. See FAQ for details.

What if I don't want to pay the Perplexity API cost? /seo-research has a manual paste fallback. Print the prompt → copy into Perplexity web UI → paste response back. Same end result, just slower.

Contributing

PRs welcome. Highest-value contributions right now:

• WordPress publisher (.claude/commands/seo-publish.md port to WP REST API)
• Contentful / Webflow / Ghost publishers
• DataForSEO MCP integration in /seo-research
• Localization workflow (one source draft → multi-language)

See CONTRIBUTING.md for guidelines. Open a Discussion before large changes.

License

MIT. See LICENSE. Use it, fork it, ship it. Stars and PRs appreciated.

Built with Claude Code by Viren Inaniyan at Asva AI / Tru Commerce.