Sophon

Sophon is an open-source programmatic SEO toolkit. It turns a seed keyword or entity list into framework-specific landing pages, sitemaps, schema markup, OG/Twitter cards, internal linking data, and optional AI-enriched content.

Install from npm:

npm install @sophonn/sophon

Framework adapters:

Who This Is For

Indie hackers 🚀

Go from a niche keyword to dozens of long-tail landing pages without building a custom pipeline from scratch.

Agencies 🤝

Standardize pSEO execution across client projects with a repeatable discovery-to-publish workflow.

Enterprise teams 🏢

Use Sophon as a controlled scaffolding layer for large SEO surfaces, while keeping routing, data, and review workflows inside your existing codebase.

Core Workflow ⚙️

1. Discover 🔎 — normalize entities from a CSV or seed keyword into data/entities.json
2. Propose 🧭 — generate intent-aware entity suggestions with priority, confidence, and action recommendations
3. Generate 🧱 — scaffold one page per entity with intent-specific sections, OG/Twitter cards, canonical, and YMYL warnings baked in
4. Technical 🛠️ — emit sitemap.xml, robots.txt, JSON-LD schema, internal link graph, and hreflang scaffold
5. Enrich ✨ — use Claude to fill TODO sections with grounded, structured content
6. Score 📊 — check entity health (metadata completeness, intent confidence, slug quality) with A-F grades
7. Audit ✅ — scan existing SEO implementation and get a weighted score with letter grade
8. Optimize 🚀 — pull real GSC performance data, detect underperforming pages, and generate prioritized recommendations

Repository Structure 📁

.
├── AGENT.md
├── README.md
├── src/
│   ├── adapters/
│   ├── cli.ts
│   ├── core/
│   ├── index.ts
│   └── types.ts
├── tests/
├── source/skills/
├── package.json
├── tsconfig.json
└── tsup.config.ts

Quickstart ⚡

1. Install 📦

npm install @sophonn/sophon

2. Run the onboarding interview 🧠

Ask your AI agent to run sophon teach. It will walk through three groups of questions — project basics, content strategy, technical setup — and save answers to .sophon.md. Every skill references this file before running.

3. Discover entities 🔍

From a seed keyword:

npx @sophonn/sophon discover --seed "best payroll software"

With custom expansion patterns:

npx @sophonn/sophon discover \
  --seed "best payroll software" \
  --pattern "{seed} alternatives" \
  --pattern "{seed} pricing" \
  --pattern "best {seed}"

From a CSV file:

npx @sophonn/sophon discover --csv ./input/entities.csv

Writes normalized entities to data/entities.json by default. Extra CSV columns are preserved in metadata.attributes for downstream use.

3.5 Propose entities (review-first) 🧭

npx @sophonn/sophon propose --seed "best payroll software"

This generates data/proposed-entities.json with intent-aware proposed entities (commercial, comparison, segmented, informational), plus priority, confidence, and recommended action (keep or review).

Use --limit to cap the list size and --propose-output to customize output path.

4. Generate pages 🧩

npx @sophonn/sophon generate --framework nextjs

One file per entity. Duplicate slugs are skipped with a warning. Pages with YMYL keywords (health, legal, financial) get editorial warnings. Thin pages are flagged as TODOs.

Intent-aware page layouts 🧠 — Sophon detects entity intent from its name and generates different TODO section scaffolds:

All pages include:

• OG and Twitter card meta tags 🏷️
• Canonical URL 🔗
• Generated comment block reminding editors not to invent facts 📝

5. Generate technical SEO assets 🗺️

npx @sophonn/sophon technical --site https://example.com

Outputs:

6. Enrich with AI 🤖

npx @sophonn/sophon enrich

Requires ANTHROPIC_API_KEY. Calls Claude to generate structured content per entity: intro, sections, FAQs, comparisons. Uses TODO markers instead of invented content when data is missing. Writes one JSON file per entity to data/enriched/.

7. Run the full pipeline 🏁

npx @sophonn/sophon run --seed "best payroll software" --framework nextjs --site https://example.com

Per-step output overrides:

npx @sophonn/sophon run \
  --seed "best payroll software" \
  --framework sveltekit \
  --site https://example.com \
  --discover-output ./data/payroll-entities.json \
  --generate-output ./src/routes \
  --technical-output ./static \
  --enrich-output ./data/enriched

8. Audit existing SEO setup ✅

npx @sophonn/sophon audit

Checks 8 SEO implementations with weighted scoring:

Reports a normalized 0-100 score with letter grade (A/B/C/D/F).

9. Score entity health 📊

npx @sophonn/sophon score

Evaluates each entity across 7 checks (title, description, tags, attributes, slug quality, intent confidence, name specificity) and assigns a score out of 100 with a letter grade. Low-scoring entities are flagged for attention. Writes results to data/scores.json.

10. Optimize with GSC data 🚀

npx @sophonn/sophon optimize --site https://example.com

Pulls real performance data from Google Search Console, maps it to your Sophon entities, and generates a prioritized optimization report.

Detects issues:

Each entity gets an optimization score (0–100) and priority level (critical/high/medium/low). Recommendations are typed as meta, content, structure, or linking with actionable steps and reasoning.

With --auto-fix, Sophon inserts [OPTIMIZE] TODO markers into enriched content files — it never overwrites existing content or modifies non-Sophon files.

Writes results to data/optimization-report.json.

Safeproof behavior (skip if already implemented) 🛡️

Generation and technical commands now skip existing non-Sophon files by default and print a reminder that something is already in place.

Use --force only when you intentionally want to overwrite:

npx @sophonn/sophon generate --framework nextjs --force
npx @sophonn/sophon technical --site https://example.com --force

Programmatic API 🧪

import {
  discover,
  propose,
  generate,
  technical,
  enrich,
  audit,
  scoreEntities,
  optimize,
  classifyIntent,
} from "@sophonn/sophon";

const proposed = propose({ seed: "best payroll software", limit: 30 });

const result = await discover({ seed: "best payroll software" });

await generate({
  entities: result.entities,
  framework: "nextjs",
  output: "app",
});

await technical({
  entities: result.entities,
  site: "https://example.com",
  output: "public",
});

await enrich({
  entities: result.entities,
  output: "data/enriched",
});

const auditResult = await audit();
// auditResult.score, auditResult.grade, auditResult.checks

const scores = scoreEntities(result.entities);
// scores.averageScore, scores.averageGrade, scores.entities

const intent = classifyIntent("best payroll software pricing");
// intent.intent → "commercial", intent.confidence → 0.9

const report = await optimize({
  site: "https://example.com",
  entities: result.entities,
});
// report.summary, report.entities[0].recommendations

Agent Skills 🤖

Sophon ships a multi-provider agent skill system. Skills are defined once in source/skills/ and distributed to:

• .claude/ — Claude Code 💬
• .agents/ — VS Code Copilot agent mode 🧭
• .cursor/ — Cursor 🖱️
• .codex/ — Codex CLI ⌨️

Six skills: sophon (master context + teach), discover, generate, technical, enrich, run.

Plus the optimize skill for GSC-powered performance analysis and recommendations.

Typical prompt:

Use Sophon to add a programmatic SEO surface to this Next.js app for the niche "best employee scheduling software". Discover entities, generate the route scaffold, and wire the technical SEO outputs into this codebase.

Current Scope 📌

What Sophon does ✅:

• Entity ingestion from CSV or seed keyword
• Intent-aware entity proposal with priority and confidence scoring
• Multi-framework page scaffolding (Next.js, Astro, Nuxt 3, SvelteKit, Remix)
• Intent-aware page layouts — different TODO sections per intent (commercial, comparison, segmented, informational)
• OG + Twitter card meta tags on every generated page
• Canonical URL per page
• YMYL detection with editorial warnings
• Duplicate slug detection (skips with warning)
• sitemap.xml with lastmod, changefreq, priority
• robots.txt with sitemap pointer
• JSON-LD schema inference (WebPage, LocalBusiness, Product, SoftwareApplication)
• Internal link graph scored by shared tags and seed keyword
• Hreflang scaffold for multilingual setups
• AI content enrichment via Claude (no-hallucination prompt, TODO markers for missing data)
• Weighted SEO audit with 0-100 score and A-F grade
• Entity health scoring (metadata completeness, intent confidence, slug quality)
• GSC-powered optimization — fetch real performance data, detect underperformance, generate typed recommendations
• Auto-fix system — safely insert optimization TODOs into enriched content
• Shared intent classification engine reusable via programmatic API
• sophon teach onboarding flow that writes project context to .sophon.md
• Multi-provider agent skill system (Claude, Cursor, Codex, VS Code)
• Output path traversal protection (all --output flags validated against cwd)
• XSS-safe template hydration (</> escaped in entity values)
• 173 tests across 13 test files (vitest)

What Sophon does not do ❌:

• OG image generation or responsive image handling
• Actual hreflang implementation (scaffold only — wire it yourself)
• Multilingual content or i18n routing
• Live SERP scraping or real-time data fetching
• Keyword research (seed keywords are user-supplied)
• Deployment or CI integration
• Analytics or tracking tag injection
• Core Web Vitals or page speed optimization
• Google Analytics integration (bounce rate, session time) — planned
• CMS sync (Contentful, Sanity, etc.)
• Content freshness management or scheduled re-enrichment
• A/B testing or variant generation

What Is Not Part of This pSEO Yet (Coming Soon) ⏳

These capabilities are intentionally out of scope today and planned for upcoming versions:

• AI-assisted entity expansion and SERP-backed discovery providers
• OG image generation per entity
• First-class multilingual support (not just hreflang scaffold output)
• Pagination and faceted navigation handling helpers
• 404 and redirect helpers (including 301 mapping outputs)
• CMS connectors (Contentful, Sanity)
• Content freshness workflows and scheduled re-enrichment
• Performance-focused generation defaults for Core Web Vitals

Design Principles 🧭

• Framework agnostic core, Next.js first output 🧩
• Minimal config to get started 🪶
• Agent-native workflow 🤖
• Built for scale across hundreds or thousands of pages 📈

Roadmap 🛣️

1. Pluggable entity discovery providers (SERP APIs, AI expansion)
2. OG image generation per entity
3. Niche-aware schema presets replacing heuristic inference
4. Review and approval workflow before publishing
5. Content freshness management and scheduled re-enrichment (sophon refresh)
6. CMS connector (Contentful, Sanity)
7. Learning loop — GSC/GA4 data feedback for entity prioritization
8. A/B testing and variant generation per intent