Claude Code Skills for Technical SEO (2026 Guide)

Screaming Frog tells you a page has a redirect chain. Then you open a ticket, wait for a developer, and check again next sprint.

Claude Code works differently. It runs inside your repository, so the same session that finds the chain can open next.config.js, collapse three hops into one, and show you the diff before you commit. That is the whole case for Claude Code skills for technical SEO, and also the source of every risk.

Start here: install AgriciDaniel’s free claude-seo plugin for its seo-technical skill (/plugin marketplace add AgriciDaniel/claude-seo, then /plugin install claude-seo@agricidaniel-claude-seo), add a Google Search Console or DataForSEO MCP server for real data, and keep Screaming Frog for full-site crawls. The other nine picks, paid options and a fix-in-the-codebase workflow follow below.

This guide covers the skills worth installing in November 2026, free and paid, with star counts and push dates pulled from the GitHub API on 1 November 2026. Prices were checked on each live product page the same day. Where a price was not shown, we say so.

What are Claude Code skills for technical SEO?

Claude Code skills for technical SEO are folders containing a SKILL.md file that teach Claude Code a repeatable audit or fix workflow: checking robots.txt, sitemaps, canonicals, redirects, Core Web Vitals, JavaScript rendering and server logs. Claude loads a skill when your prompt matches its description, then uses its terminal, file and web tools to run the checks.

A skill is plain Markdown with YAML frontmatter, plus optional scripts and reference files. It lives in one of three places:

Location	Scope	Typical use
`~/.claude/skills/<name>/SKILL.md`	Every project on your machine	Your personal audit checklist
`.claude/skills/<name>/SKILL.md`	One repository, shared through git	A team’s site-specific rules
Inside a plugin	Installed via `/plugin` marketplaces	Bundles of skills plus agents and hooks

You install them one of two ways. The open skills CLI copies a skill from a GitHub repo: npx skills add owner/repo. Plugins go through Claude Code’s marketplace commands: /plugin marketplace add owner/repo then /plugin install name@marketplace. If you are new to the format, start with what Claude Code skills are and how to install Claude Code skills.

One point matters more than the format. A skill does not add new powers to Claude. It adds a procedure. Claude Code can already run curl, read files and edit code. A good technical SEO skill tells it what to check, in what order, what “broken” means and what evidence to show you. A bad one is a long prompt with a checklist and no thresholds.

Takeaway: treat a technical SEO skill as a written audit method that an agent with shell access will follow, and read it before you install it.

Why technical SEO suits Claude Code better than most SEO work

Technical SEO is mostly pattern checking against known rules, and the fix usually lives in code you control. That combination fits an agent that can both read HTTP responses and edit the repository. Content strategy and link building fit less well, because they depend on judgement and relationships.

StudioHawk, an Australian SEO agency, makes the same split in its Claude SEO guide: audits are pattern recognition against a checklist, which is where Claude does well, while link building stays mostly human.

Here is how the main jobs line up:

Technical SEO job	What a skill can check	What Claude Code can fix in the repo
Crawlability	robots.txt rules, crawl depth, blocked resources, AI crawler rules (GPTBot, ClaudeBot, Google-Extended)	Edit `robots.txt` or the route that generates it
Indexability	`noindex`, canonicals, duplicates, thin pages, hreflang	Fix metadata in templates or the CMS layer
Sitemaps	Missing URLs, non-200 URLs, `lastmod` accuracy	Rewrite the sitemap generator
Redirects	Chains, loops, 302s that should be 301s	Edit `next.config.js`, `_redirects`, `.htaccess`, nginx config
Core Web Vitals	LCP, INP, CLS from PageSpeed or CrUX data	Defer scripts, split long tasks, size images
JS rendering	Whether key content is in the initial HTML	Move content to server rendering or static generation
Log files	Googlebot hits by status code and directory	Usually reporting only, with fixes routed elsewhere

The pain this removes

The slow part of technical SEO is rarely finding the issue. It is the handoff. An SEO exports a CSV, writes a ticket, a developer asks what a canonical is, the fix ships three weeks later with a typo, and nobody re-crawls. Claude Code collapses find, fix and verify into one session when you have repo access. You still review the diff. You stop being a messenger.

The honest catch

An agent that can edit your repo can also break it. A wrong Disallow: / line or a bad redirect rule can take a site out of the index. Every workflow below assumes a branch, a diff review and a staging check before merge.

Takeaway: the value is the shortened loop between finding and fixing, which only exists if Claude Code can see the code.

The 10 best Claude Code skills for technical SEO in 2026

The strongest free option is AgriciDaniel’s claude-seo plugin, which ships a dedicated seo-technical skill. Corey Haines’ marketingskills and rampstackco’s claude-skills add solid audit skills. Paid options include a $39 Contra pack and a $39 Kiss My Skills agent. Data comes from MCP servers such as DataForSEO and Google Search Console.

All GitHub numbers below were read from api.github.com on 1 November 2026. Stars measure attention, not audit quality.

1. claude-seo by AgriciDaniel (free)

Source: github.com/AgriciDaniel/claude-seo
Stars / last push: 25,381 stars, pushed 1 November 2026
Price: Free, open source
What it does: A plugin with 26 skills, including seo-technical, seo-sitemap, seo-hreflang, seo-schema, seo-audit and seo-google. The claudemarketplaces.com listing for seo-technical describes nine categories: crawlability (including AI crawler rules), indexability, security headers, URL structure and redirect chains, mobile parity, Core Web Vitals (LCP, INP, CLS), structured data, JavaScript rendering and IndexNow. It returns a scored breakdown with prioritised fixes. The listing notes it works best with DataForSEO or Google API access for live PageSpeed and CrUX data.
Install: /plugin marketplace add AgriciDaniel/claude-seo then /plugin install claude-seo@agricidaniel-claude-seo. For only the technical skill: npx -y skills add agricidaniel/claude-seo --skill seo-technical --agent claude-code.

This is the most complete free pick. It is also large, so expect it to pull in more context than a single-purpose skill.

2. marketingskills by Corey Haines (free)

Source: github.com/coreyhaines31/marketingskills
Stars / last push: 57,803 stars, pushed 31 October 2026
Price: Free
What it does: A broad marketing library. The relevant skills are seo-audit, site-architecture, schema and programmatic-seo. It is lighter on server-level checks than claude-seo and stronger on how technical issues connect to site structure.
Install: npx skills add coreyhaines31/marketingskills

Good if your team also wants copy and CRO skills from one source. See Claude Code skills for marketers.

3. rampstackco/claude-skills (free)

Source: github.com/rampstackco/claude-skills
Stars / last push: 1,013 stars, pushed 27 October 2026
Price: Free
What it does: Fifteen SEO skills, including seo-technical, seo-site-health-audit, seo-audit-orchestration and seo-traffic-diagnosis. Its technical skill documentation mentions producing a redirect map as a CSV for migrations, which is the most useful single output during a replatform.
Install: npx skills add rampstackco/claude-skills -a claude-code

4. alirezarezvani/claude-skills (free)

Source: github.com/alirezarezvani/claude-skills
Stars / last push: 29,110 stars, pushed 31 October 2026
Price: Free
What it does: A very large multi-domain library with a marketing-skill folder and an engineering folder. Useful when you want SEO checks plus front-end performance and engineering skills in one install.
Install: npx skills add alirezarezvani/claude-skills, or follow its INSTALLATION.md.

5. seomachine by TheCraigHewitt (free)

Source: github.com/TheCraigHewitt/seomachine
Stars / last push: 7,565 stars, pushed 23 October 2026
Price: Free
What it does: A Claude Code workspace aimed at content production and SEO analysis. It is content-first, so treat it as a companion to a technical skill rather than a replacement.
Install: Clone the repo and open it in Claude Code.

6. ComposioHQ/awesome-claude-skills (free directory)

Source: github.com/ComposioHQ/awesome-claude-skills
Stars / last push: 78,501 stars, pushed 18 September 2026
Price: Free
What it does: A curated list, not a skill. Use it to find niche technical skills, then read each one before installing.

7. DataForSEO MCP server (free server, paid data)

Source: github.com/dataforseo/mcp-server-typescript
Stars / last push: 258 stars, pushed 1 November 2026
Price: The server is free. Data is pay-as-you-go: the DataForSEO pricing page lists a $50 minimum top-up and On-Page API pricing “from $0.00015 / page”.
What it does: Gives Claude live crawl, Lighthouse and SERP data, which turns a skill’s checklist into measured findings.
Install: claude mcp add with the server command and your API credentials, per the repo README.

8. Google Search Console MCP (free)

Source: github.com/AminForou/mcp-gsc
Stars / last push: 2,086 stars, pushed 15 September 2026
Price: Free, uses your own Google credentials
What it does: Lets Claude query Search Console performance, URL inspection and sitemap status. URL inspection is what tells you whether Google actually indexed the page you just fixed.
Install: Follow the README to create OAuth or service-account credentials, then add it with claude mcp add.

Ahrefs also publishes an official MCP server (100 stars, last push 19 February 2026). It needs an Ahrefs plan with API access; we did not confirm which plans include it, so check with Ahrefs.

9. The Ultimate Claude SEO & AEO Skills Pack (paid, Contra)

Source: contra.com/products/fE9zVnna-the-ultimate-claude-seo-and-aeo-skills-pack
Price: $39 on the live page (shown against a struck-through $79), checked 1 November 2026
What it does: 12 skill files covering SEO auditing, schema, internal linking, site architecture and AI search visibility, plus a Webflow programmatic SEO bonus skill and a setup guide. The listing says lifetime updates are included. No countdown or slot-limit messaging on the page.
Install: Download the .md files and place each in its own folder under ~/.claude/skills/.

Note: the page says “2 Webflow-specific bonus skills” in one place and lists one. Ask the seller before buying if that matters to you.

10. Robert, AI Technical SEO Agent (paid, Kiss My Skills)

Source: kissmyskills.com/products/robert-ai-technical-seo-agent
Price: $39 one-time for the file, or included in the Library subscription at $9/month or $59/year, which has a 7-day free trial with no card. The page lists a 30-day money-back guarantee.
What it does: A technical audit that returns findings with severity ratings, impact notes, fix steps for WordPress, Shopify or Webflow, a verification step per fix, a quick-wins list and a 30-day plan.
Install: Download the skill file and place it under ~/.claude/skills/<name>/SKILL.md.

The same store sells a $19 Core Web Vitals and site speed skill (“Clementine”), listed on its SEO skills roundup.

What we left out

We skipped a bundle advertising “1451 Claude Skills” with an “Only 20 slots left at this price” banner, and a pack whose price rises “after the first 100 customers”. Digital files do not run out. Fake scarcity is a reason to doubt the rest of the listing.

Free vs paid Claude Code skills for technical SEO

Free GitHub skills cover more technical checks and update more often, and you can read every line before installing. Paid packs mainly sell curation, platform-specific fix steps and support. For most technical work, a free skill plus a paid data source (DataForSEO, Ahrefs or Semrush) beats a paid skill with no data.

Skill or tool	Free or paid	Price (checked 1 Nov 2026)	Best for	Install
claude-seo (`seo-technical`)	Free	$0	Full technical audits	`/plugin` marketplace or `npx skills add`
marketingskills	Free	$0	Audits tied to site architecture	`npx skills add`
rampstackco/claude-skills	Free	$0	Migrations and redirect maps	`npx skills add`
alirezarezvani/claude-skills	Free	$0	SEO plus engineering skills	`npx skills add`
seomachine	Free	$0	Content-side SEO workspace	Clone repo
GSC MCP (mcp-gsc)	Free	$0	Index status and URL inspection	`claude mcp add`
DataForSEO MCP	Paid data	$50 minimum top-up, pay-as-you-go	Live crawl and Lighthouse data	`claude mcp add`
Ahrefs MCP	Paid data	Pricing not confirmed for API access	Backlink and site audit data	`claude mcp add`
Contra SEO & AEO pack	Paid	$39	Webflow sites, bundled set	Manual copy
Kiss My Skills “Robert”	Paid	$39, or $9/month library	WordPress, Shopify, Webflow fix steps	Manual copy

Two practical rules follow from that table:

Pay for data before you pay for prompts. A skill can only judge INP if something measures INP. PageSpeed Insights has a free API; CrUX and DataForSEO add scale.
Read paid files before you rely on them. You cannot preview most paid skills. The money-back terms matter more than the feature list.

Claude Code vs Screaming Frog vs Sitebulb

Screaming Frog and Sitebulb are purpose-built crawlers that render and map an entire site reliably. Claude Code is not a crawler; it is an agent that can interpret crawl output and then fix the code. The best setup in 2026 uses a crawler for discovery and Claude Code for diagnosis and repair.

	Screaming Frog SEO Spider	Sitebulb	Claude Code + skills
Price	Free up to 500 URLs; £199/year per licence (pricing page)	Desktop Lite and Pro priced monthly per user (amounts not shown in our check); Cloud from £95/month (pricing page)	Free skills; you pay for Claude usage and any data APIs
Crawl scale	Desktop crawler, limited by your machine	Lite 10,000 URLs per audit; Pro 500,000	Small samples via `curl`; large crawls need an external crawler or API
JS rendering	Chromium rendering mode	Chromium rendering	Can compare raw HTML and rendered DOM with a headless browser, page by page
Explanations	Issue descriptions	Prioritised hints with explanations	Explains each issue in the context of your actual code
Fixes	None, exports only	None, exports only	Edits the repo and shows a diff
Log files	Separate Log File Analyser product	Not its focus	Parses logs with scripts you can read
Repeatability	Saved configs, scheduling	Scheduled audits	Skill files in git, run on demand or in CI

Where the crawlers still win: completeness. A 200,000-URL site needs a real crawler with queueing, politeness limits and a stable rendering engine. Asking Claude to curl its way through that is slow, expensive and incomplete.

Where Claude Code wins: everything after the export. Feed it the Screaming Frog “Redirect chains” CSV and it can find the rule that creates each chain in your config and rewrite it. Sitebulb’s hints are good explanations; Claude Code turns the explanation into a commit.

Takeaway: keep your crawler. Add Claude Code at the point where you used to write a ticket.

How to run a technical SEO audit and fix it in the codebase

Install a technical skill, connect Search Console and a performance data source, run the audit on a branch, ask Claude to fix one issue class at a time, review each diff, deploy to staging, and verify with live HTTP checks and URL inspection before merging.

Here is the workflow, step by step:

Branch first. git checkout -b seo/technical-fixes. Never let an agent edit main for SEO changes.
Install the skill. For example, npx -y skills add agricidaniel/claude-seo --skill seo-technical --agent claude-code.
Connect data. Add the GSC MCP server and either DataForSEO or the PageSpeed Insights API.
Run the audit. Prompt: “Run a technical SEO audit of https://example.com. List issues by severity with the URL and evidence for each.”
Import crawler output if you have it. Drop the Screaming Frog or Sitebulb CSV into the repo and ask Claude to merge it with the audit.
Fix one class at a time. “Fix every redirect chain listed. Show me the config change and which URLs it affects.” Smaller diffs are easier to review.
Verify locally. Ask Claude to run curl -sIL against the dev server for each changed URL and show the hop count.
Stage and re-check. Deploy to staging, re-run the skill, compare.
Merge and inspect. After production deploy, use URL inspection through the GSC MCP to request recrawl of the key URLs.

Crawlability and robots.txt

Ask Claude to fetch /robots.txt, list every Disallow, and test your ten most important URLs against those rules. Then ask it to find where the file is generated. In Next.js that is often app/robots.ts; in WordPress it may be a plugin setting, which Claude cannot change from the repo. Decide on AI crawler policy (GPTBot, ClaudeBot, Google-Extended) as a business choice, then have Claude write it. Related: Claude Code skills for GEO.

Indexability and canonicals

The common bug is a template that sets the canonical to the homepage, or to a URL with tracking parameters. Claude can grep the layout files for rel="canonical" and show you exactly where the value comes from. That is a five-minute fix that crawlers can only report.

XML sitemaps

Ask Claude to fetch the sitemap, request every URL, and flag anything that is not a 200, is redirected, or is canonicalised elsewhere. Then fix the generator so excluded pages stay out. The seo-sitemap skill in claude-seo covers this.

Redirects

Redirect rules live in many places: next.config.js, vercel.json, Netlify _redirects, .htaccess, nginx, Cloudflare. Tell Claude which one you use. During migrations, have it build a full old-to-new redirect map as a CSV, test each row, and only then write the rules.

Core Web Vitals and INP

Interaction to Next Paint replaced First Input Delay as a Core Web Vital in March 2024. Google’s web.dev guidance defines a good INP as 200 milliseconds or less at the 75th percentile of page loads. Poor INP usually comes from long main-thread tasks: heavy event handlers, hydration, third-party scripts. Claude Code can find the handler, break up the work (for example yielding with scheduler.yield() or setTimeout), defer non-critical scripts and lazy-load widgets. It cannot measure field INP itself, so pair it with CrUX or PageSpeed data. Lab data alone will mislead you on INP.

JavaScript rendering

The test is simple: does the important content appear in the raw HTML? Ask Claude to compare curl output with the rendered DOM from a headless browser and list headings, links and body text that only appear after JavaScript runs. If product descriptions or internal links are client-only, the fix is moving them to server rendering or static generation, which is a code change Claude can make and you must test.

Log file analysis

Server logs show what Googlebot actually requested, which no crawler can tell you. Ask Claude to write a short script that parses your access logs, filters verified Googlebot requests (check reverse DNS, since user agents are easy to fake), and groups hits by status code and directory. Review the script before running it on production logs, and strip personal data such as IP addresses if you share the output.

Takeaway: each issue class has a place in the code. The skill finds the issue; your prompt should always ask where in the repo it comes from.

Common mistakes with Claude Code technical SEO skills

The big mistakes are letting the agent edit production directly, trusting lab data for INP, installing skills without reading them, assuming curl checks equal a full crawl, and shipping robots.txt changes without testing important URLs against them.

Mistake	Why it hurts	What to do instead
Editing `main` directly	One bad robots rule can deindex the site	Branch, diff review, staging
Installing unread skills	Skills run with your shell and file access	Read `SKILL.md` and any scripts first
Treating sample checks as a crawl	Misses orphan pages and deep issues	Use a crawler for discovery
Judging INP from Lighthouse	Lighthouse reports Total Blocking Time, not field INP	Use CrUX or RUM data
Fixing everything in one diff	Impossible to review or roll back	One issue class per commit
Skipping verification	Fixes often fail silently on CDN or cache layers	Re-check live headers after deploy
Ignoring CMS limits	Some settings live outside the repo	Note which fixes need CMS access

The security point deserves its own line. Community registries such as claudskills.com and skills.sh host thousands of skills from unknown authors. A skill can include scripts. Read them, the same way you would read an install script before piping it to bash. claude-seo’s own README tells you to download and inspect install.sh before running it, which is the right habit.

Who should use Claude Code for technical SEO, and who should not

Use it if you or your team can access the site’s repository and review code diffs. Skip the fix-in-codebase workflow if your site is a closed SaaS builder, if you cannot review code, or if a separate engineering team owns deployment and will not accept agent-made changes.

Good fits:

In-house SEOs at companies with a modern codebase (Next.js, Astro, Nuxt, Rails, Django)
Technical SEO consultants who are allowed to open pull requests
Developers who own their own marketing sites
Agencies building repeatable audit methods as skills in a shared repo. See how to build a Claude Code skill.

Weaker fits:

Sites on Wix or Squarespace, where most technical settings are not in code you can edit
Teams where nobody can read a diff. The audit part still helps, but fixes need a developer
Very large sites, where crawl and log volume call for dedicated tools first

For on-page and content work, see Claude Code skills for SEO, Claude Code skills for SEO audits, schema markup skills and internal linking skills. Running large template-driven sites? Read Claude Code skills for programmatic SEO.

FAQ

Can Claude Code do technical SEO?

Yes, within limits. It can fetch pages, read headers, parse robots.txt and sitemaps, compare raw and rendered HTML, analyse logs with scripts and edit your code. It is not a full-site crawler, so pair it with Screaming Frog, Sitebulb or a crawl API for large sites.

Can Claude replace Screaming Frog?

Not for crawling large sites. Screaming Frog maps every URL reliably and costs £199 per year after the free 500-URL tier. Claude Code is better at explaining and fixing what the crawl finds. Most teams will use both.

What is the best free Claude Code skill for technical SEO?

AgriciDaniel’s claude-seo has the most complete free seo-technical skill as of November 2026, covering crawlability, indexability, redirects, Core Web Vitals, JS rendering and IndexNow. rampstackco/claude-skills is a good second choice for migrations.

Are paid Claude SEO skill packs worth it?

Sometimes. The two we checked cost $39 each and add platform-specific fix steps for Webflow, WordPress or Shopify. You cannot read them before buying, so check refund terms. For most technical work, spending on data APIs gives better results.

How do I install a technical SEO skill in Claude Code?

Run npx skills add owner/repo, or use /plugin marketplace add owner/repo and /plugin install. For a single downloaded file, create ~/.claude/skills/<name>/SKILL.md and paste it in. Claude Code watches the skills folders, so no restart is needed. If ~/.claude/skills/ did not exist when the session started, run /reload-skills.

Does Claude Code check Core Web Vitals?

Only with data. A skill can call the PageSpeed Insights API, CrUX or DataForSEO to get LCP, INP and CLS. Without a data source it can only inspect code for likely causes.

What is a good INP score?

200 milliseconds or less at the 75th percentile is good, per Google’s web.dev guidance. Between 200 and 500 ms needs improvement. Above 500 ms is poor.

Can Claude Code fix redirect chains?

Yes, if the redirect rules are in your repository or a config file it can read. Tell it which layer handles redirects, ask it to collapse chains to a single hop, and verify each URL with curl -sIL after deploying.

Can Claude Code check if JavaScript content is indexable?

It can compare raw HTML with a headless-browser render and list what only appears after JavaScript. That shows risk. Confirm what Google sees with Search Console’s URL inspection, which the GSC MCP server exposes.

Can Claude analyse server log files for SEO?

Yes. Ask it to write a parsing script, verify Googlebot by reverse DNS, and summarise hits by status code, directory and date. Read the script first and remove personal data before sharing results.

Is it safe to let Claude Code edit robots.txt?

Only with review. Work on a branch, ask Claude to test your key URLs against the new rules, and deploy to staging first. A single wrong Disallow line can block a whole site.

Do Claude Code skills work with Screaming Frog exports?

Yes. Put the CSV in your project and ask Claude to read it. It can group issues, find their source in templates or config, and propose fixes.

Which MCP servers help with technical SEO?

Google Search Console (index status, URL inspection), DataForSEO (crawl and Lighthouse data, pay-as-you-go from a $50 top-up), and Ahrefs (site audit and backlink data, needs an Ahrefs plan). Semrush data can also be connected; check its current API terms.

Does Claude Code need repository access for technical SEO?

For audits, no. It can test any public URL. For fixes, yes. The fix-in-codebase workflow is the main advantage over desktop crawlers, and it requires the repo.

What is the difference between a skill and an MCP server?

A skill is a written procedure Claude follows. An MCP server is a connection that gives Claude new data or actions, such as Search Console queries. Technical SEO works best with both: the skill decides what to check, the server supplies the measurements.

Sources and methodology

GitHub stars and last-push dates: api.github.com/repos/<owner>/<repo>, read 1 November 2026.
Skill contents: repo directory listings and READMEs; seo-technical scope from claudemarketplaces.com.
Prices: Contra pack, Kiss My Skills, DataForSEO, Screaming Frog, Sitebulb, all checked 1 November 2026. Prices change; check before buying.
INP thresholds: web.dev/articles/inp.
Practitioner framing: StudioHawk.
No skill in this list was benchmarked by us. Rankings reflect documented scope, maintenance activity and price, not measured audit accuracy.

Your next step

Pick one site you can deploy. Install claude-seo’s seo-technical skill, connect Search Console, and run the audit on a branch. Fix one issue class, review the diff, verify it live. If that loop works for redirects, it will work for canonicals, sitemaps and INP. Then write your own rules into a project skill so the next audit starts where this one ended. Build your own Claude Code skill.

Claude Code Skills for Technical SEO: 10 Free and Paid Picks (2026)