SEO Content Pipeline

4 stages. Keywords to drafts. No CMS required.

Keyword research → Competitor scraping → SEO gap analysis → 1,500+ word article drafts
_{Runs in any AI coding agent with tool-calling. Outputs local files — zero CMS dependencies. Browser Use Cloud SDK for AI-powered article extraction; DuckDuckGo fallback for URL discovery.}

---

Pipeline

flowchart LR
  A["keyword<br>research"] --> B["content<br>collection"] --> C["SEO<br>analysis"] --> D["article<br>drafting"]

Loading

---

Install

No bootstrap scripts, no setup commands. Each agent has a built-in path.

Hermes Agent

Install as a plugin:

hermes plugins install attaxa/seo-pipeline --enable

Or load the pipeline skill in-session:

skill_view(name='seo-pipeline-llm')

Then in chat say "run the SEO pipeline for [topic]".

Schedule as a recurring cron job:

cronjob action=create schedule="0 9 * * 1" name="weekly-seo" \
  prompt="Run the SEO content pipeline for [topic]. Execute all 4 stages." \
  skills='["seo-pipeline-llm"]'

Claude Code

Reads AGENTS.md from the project root automatically — no setup needed.

claude AGENTS.md -- "run stage 1 keyword research for [topic]"

Or from an open repo session:

"run the SEO pipeline for [topic]"

Codex

Codex reads AGENTS.md automatically from the repo root. Clone and go:

codex
# Then: "run the SEO pipeline for [topic]"

OpenCode

Configured via opencode.json in the repo:

{ "agent": { "instructions": "AGENTS.md" } }

OpenCode loads pipeline instructions every session.

Copilot CLI

Reads .github/copilot-instructions.md from the project root. Clone and go.

For global availability, copy the file:

cp .github/copilot-instructions.md ~/.copilot/instructions/

Cursor

.cursor/rules/seo-pipeline.mdc ships in-repo — Cursor loads it automatically when you open the project.

For global install:

cp .cursor/rules/seo-pipeline.mdc ~/.cursor/rules/

Windsurf

The repo ships .windsurf/rules/seo-pipeline.md — Windsurf loads it from the project root automatically.

For global install:

cp .windsurf/rules/seo-pipeline.md ~/.windsurf/rules/

Cline / Aider

Clone the repo and start a session. Cline reads .clinerules/seo-pipeline.md from the project root. Aider reads AGENTS.md.

Pi agent harness

pi install git:github.com/attaxa/seo-pipeline

AGENTS.md fallback

Every agent that reads project-level AGENTS.md / CLAUDE.md / .cursorrules picks up the pipeline instructions automatically. Just clone the repo and start a session:

Agent	Reads from project root
Claude Code	AGENTS.md, CLAUDE.md
Cursor	.cursor/rules/, .cursorrules, AGENTS.md
Windsurf	.windsurf/rules/, AGENTS.md
Cline	.clinerules/, AGENTS.md
Aider	AGENTS.md
Codex	AGENTS.md
OpenCode	AGENTS.md, opencode.json
Copilot CLI	.github/copilot-instructions.md

Uninstall

Host	Command
Hermes	Remove the skill reference from your session or cron job
Claude Code	Stop referencing AGENTS.md — or remove the file from your local clone
Cursor	rm ~/.cursor/rules/seo-pipeline* (if globally installed)
Windsurf	rm ~/.windsurf/rules/seo-pipeline* (if globally installed)
Any	Delete the copied rule or skill files

---

Configuration

Variable	Default	Description
SEO_PIPELINE_DIR	./pipeline_data/	Directory for pipeline artifacts
BROWSER_USE_API_KEY	—	Browser Use Cloud API key (get one at https://cloud.browser-use.com/settings?tab=api-keys)
BROWSER_USE_PROXY_COUNTRY	us	Residential proxy country code (ISO 3166-1 alpha-2). Set to none to disable
BROWSER_USE_CUSTOM_PROXY_HOST	—	Custom proxy host. Overrides residential proxy when set
BROWSER_USE_CUSTOM_PROXY_PORT	8080	Custom proxy port
BROWSER_USE_CUSTOM_PROXY_USERNAME	—	Custom proxy username (optional)
BROWSER_USE_CUSTOM_PROXY_PASSWORD	—	Custom proxy password (optional)

Set via .env or export. Run python scripts/setup.py for interactive configuration with redacted key input.

---

Commands

Command	What it does
python scripts/setup.py	Interactive Browser Use setup + .env (redacted key input)
python -m seo_pipeline --from-json keywords.json	Scrape articles for all keywords
python -m seo_pipeline "keyword"	Scrape articles for one keyword
"run the SEO pipeline for [topic]"	In-agent: full 4-stage execution

---

How It Works

See AGENTS.md for the full pipeline description, quality checklist, and output structure.

---

Requirements

• Python 3.11+
• An LLM agent with tool-calling (web search, file I/O, terminal)
• BROWSER_USE_API_KEY for full article extraction
• (DuckDuckGo fallback needs no API key but only discovers URLs)

---

Agent Portability

Agent	Config file	Setup
Hermes	plugin.yaml + __init__.py + skills/ + commands/	hermes plugins install attaxa/seo-pipeline --enable
Claude Code	AGENTS.md + .claude-plugin/	Clone repo (marketplace: .claude-plugin/marketplace.json)
Cursor	.cursor/rules/	Ships in-repo
Windsurf	.windsurf/rules/	Ships in-repo
Copilot CLI	.github/copilot-instructions.md + hooks/copilot-hooks.json	Ships in-repo
Codex	AGENTS.md + .codex-plugin/	Clone repo
OpenCode	opencode.json + .opencode/	Clone repo
Devin	.devin-plugin/	Ships in-repo
Cline	.clinerules/	Ships in-repo
Aider	AGENTS.md	Clone repo
Pi	pi-extension/	pi install git:github.com/attaxa/seo-pipeline
Gemini	gemini-extension.json	Register in Gemini Studio

---

Contributors

---

License

MIT. See LICENSE.