claude-seo: A Free SEO Audit Plugin for Claude Code, Explained

A Claude Code plugin that audits your website for search and for AI answer engines is trending on GitHub. AgriciDaniel/claude-seo has passed 20,000 stars, gained about 179 stars in a single day this week, and describes itself as a "universal SEO skill for Claude Code" with 26 sub-skills and 19 sub-agents. The repo covers technical SEO, E-E-A-T, schema, GEO and AEO, and "agent readiness" checks such as llms.txt and WebMCP.

This guide explains what it does, how to install it, what a run costs in Claude usage, how it handles your data, and where to be skeptical. It is a tool walkthrough, not an endorsement: the performance numbers in the README come from the author's own site.

TL;DR: questions people ask first

How claude-seo is built

The plugin follows the pattern explainx.ai has covered in what agent skills are: a set of skills that Claude Code loads on demand, plus sub-agents that run in parallel. The README says an audit "detects the type of business" (SaaS, local, e-commerce, publisher or agency), then up to 17 specialist agents audit the site at the same time and the findings are scored into a plan. Each recommendation says what it is based on and how you will know it worked.

The agent table in the README shows the split of work:

You can change an agent's model in its file under the agents/ folder, which is the main lever for cost.

A checklist clipboard representing a prioritized website SEO audit plan

The commands worth knowing

The README lists 30 core and extension commands in six groups. These are the ones most people will use first:

The agent readiness command is the most current. It checks whether AI agents can use your site, which ties into the standards explainx.ai has covered in the llms.txt specification guide and in Sodium WebMCP, which turns website features into agent tools.

How to install it

From inside Claude Code, the README gives three commands:

/plugin marketplace add AgriciDaniel/claude-seo
/plugin install claude-seo@agricidaniel-claude-seo
/seo setup

The plugin mechanism is documented in the Claude Code plugins docs. For a manual install, the README shows a git clone followed by install.sh on macOS and Linux, or install.ps1 on Windows. It deliberately avoids a pipe-to-shell install and tells you to read the script first. That is good practice for any plugin that installs a Python environment and a browser.

Then run a first audit:

claude
/seo audit https://example.com
/seo page https://example.com/pricing
/seo geo https://example.com

What does an audit cost?

The README says the plugin is free, with no per-site pricing and no tracking. The real cost is your Claude usage. Five of the agents run on Opus and the rest on Sonnet, and the README itself warns that a full /seo audit uses more of your plan than a single-page check. If you are on a metered or limited plan, start with /seo page and lower the model on agents you do not need. For how usage limits behave in Claude Code, see explainx.ai's Claude Code commands reference.

What does it do with your data?

The README's FAQ says that because it runs inside Claude Code, the pages and data it analyzes are sent to Claude, "the same as anything else you work on in Claude Code." Reports are saved as files on your machine. Some checks read public sources such as Wikipedia, Reddit and Common Crawl. Optional extensions (image generation, SE Ranking, Profound, Bing Webmaster, Unlighthouse) connect to third parties only if you set them up.

For client work, that matters: an agency auditing a client's staging site is sending that content to a model provider. Check your contracts before auditing anything non-public.

The grain of salt: self-reported results

The README says the project's own site, audited with the tool "from day one," earned 10.8K Search Console clicks in its first 12 weeks and a 99 mobile PageSpeed score. That is a single self-reported case. It does not show that the tool's recommendations raise rankings elsewhere. The comparison table in the README (hours by hand versus 10 to 15 minutes) is labeled as typical market prices for illustration.

Three honest limits from the README itself:

1. Without Google API keys, Core Web Vitals are lab estimates and indexing is inferred.
2. Pages that load key content after scrolling or clicking can give noisy results.
3. Crawlers such as Screaming Frog crawl faster and deeper, and big suites have their own backlink data.

A language model reading your pages can also misjudge E-E-A-T or invent a reason a fix will work. Verify any recommendation against Google's own documentation before shipping it.

How it relates to GEO and AI search

The plugin leans into GEO, the practice of getting cited by AI answers. Its README says Google states that optimizing for AI search is still SEO, so it starts with indexing, speed and clarity, then scores how quotable pages are, whether AI crawlers can reach them, and the brand's presence on sites like Wikipedia, Reddit, YouTube and LinkedIn. If you want the concepts first, read explainx.ai's explainers on SEO and GEO, the GEO guide for marketers and the warning about AI slop in SEO content. Our own SEO and GEO agent skill is a lighter alternative.

What an audit report looks like

The README shows the start of a schema report for a sample page. It lists the number of JSON-LD blocks (3), the schema types found (Organization, WebSite, SoftwareApplication), two critical issues, five warnings, 18 passed checks and an overall grade of B+. The example critical issue is a content-schema mismatch: the page displays a co-founder who is missing from the founder property of the markup.

That example is useful because it shows the kind of finding a language model is good at. It compares what a human sees on the page with what the structured data says, a tedious cross-check that crawlers do not do. It is less reliable at things that need field data, such as whether a page really ranks, which is why the README points to Search Console and the Chrome UX Report for real numbers.

Follow-up prompts to try after an audit

Because the results live in Claude Code, you can keep working with them in plain English. Some prompts that fit the plugin's design:

• "Show only the critical issues from the last audit and estimate the effort for each."
• "Write the missing schema for my pricing page, then explain how to validate it."
• "Which of these findings would also help AI crawlers reach the page?"
• "Compare this page against the top three results for my target query and list gaps."

Treat the answers as hypotheses. The point of the /seo drift commands is to measure whether a change did anything, so use them.

A grey area: pro version and funnel

The README says the public version is MIT-licensed and free, while members of a paid community, AI Marketing Hub Pro, get a private version with early access to new features. The project is also part of a family of related tools (a blog writer, image generation, a ChatGPT version). That is normal for open-source marketing tools, but it explains why the repo is well polished and promoted. Judge the free version on its own output.

A sensible workflow

1. Run /seo doctor and /seo page on one URL to see the report format.
2. Connect free Google APIs only if you want field data, and keep keys on your machine.
3. Run /seo audit on a staging copy first if the site is non-public.
4. Take the top three fixes, make them by hand, and re-run /seo drift compare to confirm.
5. Never let an agent publish site-wide changes without review, and test any schema it writes with a validator.

Bottom line

claude-seo is a well-documented example of what a skill-and-sub-agent plugin can look like in Claude Code: a clear command set, parallel agents, local reports and honest limits. Its star count shows demand for audits that cover AI search. Whether its advice moves rankings for your site is something only your own before-and-after data can show.

Related reading

• What are agent skills? A complete guide
• SEO and GEO: generative engine optimization in 2026
• GEO for marketers: AI answer engines
• The llms.txt specification guide
• Sodium WebMCP: website features as agent tools
• SEO and GEO agent skill for AI search
• Claude Code commands reference

Details are accurate as of October 25, 2026. Star counts and command lists change often; check the repository.