An SEO audit plugin inside Claude Code
Maintained by Agrici Daniel, claude-seo is a Claude Code SEO plugin under the MIT license. Its core function divides website SEO audits among 25 sub-skills and 18 expert agents, dispatching work in parallel with a single command: technical SEO, E-E-A-T content quality, Schema.org, GEO/AEO (AI search optimization), local SEO, ecommerce SEO and international hreflang each have dedicated agents, whose results form a Markdown report with priority levels.
It is not a new SaaS product and requires no per-domain payment. Skills and agents are discovered automatically in skills/seo-*/ and agents/seo-*.md; when you enter /seo audit <url>, the orchestrator first detects the industry, dispatches up to 15 agents in parallel, then synthesizes their results. The architecture has three layers: directive (instructions), orchestration (coordination) and execution.
The difference from standard SEO tools is the methodology: each audit follows four stages, Perceive → Analyze → Validate → Act, and 10 principles. Every recommendation includes four things: first-principles observations, dependencies, a falsification check ("How do we know this failed?") and leading indicators to monitor progress. The full framework is in skills/seo/references/thinking-framework.md.
Perceive
Analyze
Validate
Act
Every recommendation includes a falsification check: a test explicitly stating "How do we know this failed?"
Two plugin commands or a single clone
Requirements: Python 3.10+ and Claude Code CLI. Playwright Chromium (for SPA rendering) and Google API credentials are optional; Chromium installs automatically. The fastest installation method is Claude Code's plugin marketplace: run the following two lines in the chat.
# Run inside Claude Code
/plugin marketplace add AgriciDaniel/claude-seo
/plugin install claude-seo@agricidaniel-claude-seoManual installation (Unix / macOS / Linux)
If you prefer a direct clone, run the installation script; use the PowerShell script on Windows.
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
bash claude-seo/install.shgit clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
powershell -ExecutionPolicy Bypass -File claude-seo\install.ps1/seo commands and audit areas
All features use the /seo namespace. /seo audit is the entry point: it dispatches multiple expert agents in parallel for a full-site audit; the other commands analyze one area in depth.
See docs/COMMANDS.md for all 25 commands. In practice, remembering audit, schema, geo, google and drift covers most work.
Entry · 01
/seo audit <url>
Full site audit
Dispatch multiple sub-agents in parallel for a full-site audit and synthesize an action list with priority levels after detecting the industry.
Page · 02
/seo page <url>
Single-page analysis
Inspect one URL in depth, suitable for further diagnosis of important pages after an audit.
Technical · 03
/seo technical <url>
Nine technical SEO categories
Technical SEO audits across 9 categories: indexing, crawlability, Core Web Vitals and others.
Content · 04
/seo content <url>
E-E-A-T content quality
Assess E-E-A-T against the 2025/9 Quality Rater Guidelines, including filler detection and humanizing AI writing patterns.
Structured data · 05
/seo schema <url>
Schema detection and generation
Detect, validate and generate JSON-LD for 15+ types including Organization, Article and Product, and flag deprecated schema.
AI search · 06
/seo geo <url>
GEO / AI Overviews
Assess paragraph citability (optimal 134–167 characters), question headings and entity signals using Google's AI optimization guidelines.
Local · 07
/seo local <url>
Local SEO / GBP
Audit Google Business Profile signals, NAP consistency across citation sources and review sentiment, including detection of deprecated GBP fields.
Links · 08
/seo backlinks <url>
Backlink analysis
Analyze the link profile and assess backlink sources and structure.
International · 09
/seo hreflang <url>
International SEO / i18n
Internationalization and hreflang audits, including machine translation quality checks and DMA consent mode diagnosis.
Ecommerce · 10
/seo ecommerce <url>
Ecommerce SEO audit
Ecommerce SEO and marketplace intelligence, including product structured data validation.
API · 11
/seo google [cmd]
Google API integration
Connect GSC, PageSpeed, CrUX, GA4 and Indexing API with Tier 0–3 credential levels, usable without payment.
Monitoring · 12
/seo drift <url>
Change monitoring
The three baseline / compare / history subcommands track changes in site SEO over time with SQLite snapshots.
Output · 13
/seo sitemap generate
Sitemap generation
Generate a sitemap from audit findings for submission and indexing.
Goals and corresponding commands
| Your goal | Corresponding command |
|---|---|
| First encounter with the site, requiring a comprehensive audit | /seo audit |
| Make content easier for AI Overviews to cite | /seo geo |
| Check and generate structured data (JSON-LD) | /seo schema |
| Connect GSC / PageSpeed / GA4 for measured data | /seo google |
| Track site SEO changes over time | /seo drift baseline | compare |
Ways to improve audit accuracy
All operating tips come from the official README and documentation and contain no unverified rumors. The design priorities are parallel execution, falsifiability and local-first operation.
Full-site audits and focused analysis
Run /seo audit <url> as the entry point for a full-site audit, then use /seo schema, /seo geo and other commands to examine important areas after obtaining the list with priority levels. audit automatically detects the industry and dispatches work in parallel.
Source · Official README · Quick Start
GEO mode and AI search
/seo geo assesses paragraph citability using Google's AI optimization guidelines: optimal blocks have 134–167 characters, use question headings and strengthen entity signals (Wikipedia, Reddit, YouTube, LinkedIn). Run it to target citations in AI Overviews.
Source · Official README · AI-Search Alignment
Measured data from Google APIs
/seo google uses Tier 0–3 credential levels. Even without enabling any paid services, it can obtain measured PageSpeed, Search Console, CrUX and GA4 data, replacing estimates with measured data in recommendations.
Source · Official README · Google API Integration
Site change monitoring with drift
Save a SQLite snapshot with /seo drift baseline, then use compare and history for later comparisons. This turns individual audits into a trackable time series, suitable for comparing before and after a redesign or for regular checks.
Source · Official README · Core commands
Automatic Playwright rendering for SPAs
v2 Phase A adds headless rendering: SPAs such as Next.js, React and Vue obtain rendered content through automatically installed Playwright Chromium, avoiding misdiagnosis from seeing only an empty HTML shell.
Source · Official README · Version 2 Phase A
Falsification checks for every recommendation
Every report recommendation includes four things: first-principles observations, dependencies, a falsification check and leading indicators. Read the falsification check before implementation to identify which indicators to use when verifying effectiveness.
Source · Official README · 10 methodology principles
/seo audit full-site audit workflow
This illustrative workflow runs /seo audit on an ecommerce site. The orchestrator first detects the industry, dispatches multiple expert agents in parallel and synthesizes their reports into a list with priority levels; next, /seo schema adds structured data, /seo geo checks AI search readiness, and /seo drift baseline saves a baseline.
$ You › /seo audit https://acme-shop.com
claude: claude-seo ›
Perceive → Analyze → Validate → Act
render: SPA detected as Next.js → Obtain rendered content with Playwright Chromium
hl: Prioritize(P0)
1. Product schema missing offers.priceValidUntil → Most product pages
2. LCP p75 = 4.1s(CrUX)→ hero image lacks priority
3. Category page headings are not questions → Low AI Overviews citation rate
"After the fix,Rich Results Test should report 0 errors for Product;
valid items in GSC's 'Products' enhancement report should rise within two weeks."
ok: [wrote ./reports/acme-shop/audit-2026-06-16.md]
ok: 25 sub-skill · 18 agent · Completed in 12 minutes]
$ You › /seo schema https://acme-shop.com/p/widget-x
claude: Complete Product JSON-LD generated(including offers,aggregateRating);
$ You › /seo drift baseline https://acme-shop.com
ok: [SQLite snapshot saved · Use /seo drift compare next time to compare changes]
Dispatch 15 agents in parallel and finish in 12 minutes; every recommendation includes a falsification check and leading indicators that can be verified later.
Audit structure and verification
The orchestrator first detects the industry, then dispatches expert agents by area in parallel; SPA sites first obtain rendered content through Playwright to avoid treating an empty HTML shell as complete content. Results are synthesized into a report with priority levels rather than scattered alerts.
Every recommendation includes a falsification check and leading indicators, requiring a definition of "How do we know this works?" before implementation. This replaces intuitive SEO adjustments with trackable work items that have acceptance criteria, the largest difference from standard audit tools.
Audit tool limitations
MCP extensions and live data
Core audits run locally. For more detailed data, claude-seo has 8 optional MCP extensions, each installed through extensions/[service]/install.sh with independently managed credentials.
Optional MCP extensions
1. DataForSEO provides SERP data, keyword research, backlinks and AI visibility.
2. Firecrawl provides full-site crawling and URL discovery, suitable for content inventories on large sites.
3. Banana generates AI images: OG previews, blog hero images and infographics.
4. New in v2: five new extensions, Ahrefs, SE Ranking, Profound, Bing Webmaster and Unlighthouse, add link, ranking, AI visibility and performance data.
Methodology and data quality
v2 Phase A–F substantially expands capabilities: SPA rendering, QRG-aligned content gates, deeper Core Web Vitals analysis (LCP subparts, Speculation Rules, IndexNow), Schema completeness, AI search restructuring, and local and internationalization improvements. Test coverage reaches 271 passing tests (a 5.4-fold increase over v1.9.9), including 83 SSRF and DNS rebinding bypass tests.
Further reading
① docs/COMMANDS.md: Reference for all 25 commands. ② github.com/AgriciDaniel/claude-seo: Source code, installation and architecture documentation. ③ skills/seo/references/thinking-framework.md: The full methodology framework with four stages and 10 principles.
Each audit follows Perceive → Analyze → Validate → Act, anchored in 10 principles; every recommendation can be verified.