Claude SEO: The Free Open Source Tool That Turns Claude Code Into a Full SEO Command Center
Claude SEO is a free plugin that adds a full SEO command set to Claude Code. Updated for v2.4.2, with install steps, every command, my own audit results, and the other SEO tools that connect to Claude.
Explore with AI

Table of Contents
Claude SEO is a free, open source SEO plugin for Claude Code. Type /seo audit https://yoursite.com and it audits the site, scores it from 0 to 100 and writes a prioritized action plan as local markdown files. It has grown a lot since this guide first went up in April. As of October 11, 2026 the project is at v2.4.2, with 26 subskills, 19 specialist agents and 9 optional extensions (the README counts 34 commands), and its GitHub repository has about 18,700 stars.
I rewrote this guide to match. Ahrefs now starts at $129 a month for its Lite plan, and Semrush's SEO plan is $139 a month billed monthly (see the Ahrefs and Semrush pricing pages), so a free audit tool that runs inside the terminal you already use is easy to try. I ran the current plugin scripts against this site for this refresh, and I show my own audit results below. Claude SEO is an audit tool, not a rank tracker, so further down I also cover the other SEO tools that connect to Claude today, including the ones that do track rankings.
One disclosure up front. My July audit ran on an older install (the copy on my machine is v1.7.0), and for this refresh I ran the v2.4.2 Python scripts but not the slash commands themselves, because the nested Claude Code session I used for testing could not authenticate. So the command descriptions come from the project's documentation, and the outputs I show are real script runs.

What is Claude SEO?
Claude SEO is a plugin for Claude Code, Anthropic's command line tool for AI assisted development. It was built by Agrici Daniel, who describes himself in the README as an AI workflow architect and the single maintainer. Once installed, it adds /seo commands that analyze websites for technical SEO, content quality, schema markup, Core Web Vitals, AI search readiness, local SEO and more.
The analysis runs locally. According to the project's README, the core plugin makes no third party API calls by default, although audits still fetch the URLs you point them at, and it carries no telemetry. It costs nothing beyond the Claude Code subscription you already pay for. The code is MIT licensed, and the README also describes a private community mirror with early access for members of the author's membership community, so check which repository you install from.
The official repository and how to check you have it
If you searched for "claude seo" to find the code itself, this is the project's own repository. Every figure in the table comes from the GitHub API on October 11, 2026.
| Item | Value (October 11, 2026) |
|---|---|
| Repository | github.com/AgriciDaniel/claude-seo |
| Maintainer | Agrici Daniel, GitHub account AgriciDaniel, the single maintainer named in the README |
| Licence | MIT |
| Latest release | v2.4.2, published October 4, 2026 |
| Stars and forks | 18,666 stars and 2,734 forks |
| Project website | claude-seo.md, the homepage set on the repository |
The README gives three install routes today: the plugin marketplace and a Windows PowerShell script, both covered step by step below, and this manual install for macOS and Linux:
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
bash claude-seo/install.sh
Before you run it, check two things. First, the owner in the URL must be AgriciDaniel. A copy under any other account is a fork or a mirror, and the only other source the README names is the private community mirror, AI-Marketing-Hub/claude-seo, which requires membership in the author's Pro community. On the plugin route, the marketplace name to expect is agricidaniel-claude-seo, and the plugin manifest lists AgriciDaniel as the author and claude-seo.md as the homepage.
Second, read install.sh before it runs. The v2.4.2 script, as published on October 11, 2026, checks that git is installed, clones the repository again from https://github.com/AgriciDaniel/claude-seo pinned to the v2.4.2 release tag, copies the skills into ~/.claude/skills/ and the agents into ~/.claude/agents/, then builds an isolated Python environment. If the copy you downloaded fetches from any other address, stop. If you prefer curl, the README's own option saves the script to a file, prints it for you to read and only then runs it, rather than piping it straight into bash.
Is it safe to install? A star count measures popularity, not security, and this guide is not a code audit. Claude Code's plugin security page says a plugin "can execute arbitrary code on your machine with your user privileges", and it treats every marketplace outside Anthropic's reserved names as third party, which includes this one. The points in this project's favor are public: the code is MIT licensed, every release has notes, and the changelog records its security fixes. The same page shows how to list a plugin's skills, agents, hooks and MCP servers before you install it. From the folder where you ran git clone, run claude --plugin-dir ./claude-seo plugin details claude-seo.
The current architecture, from the README:
- One orchestrator skill that routes each command to the right specialist
- 26 subskills covering technical SEO, content, schema, sitemaps, performance, images, AI search, agent readiness, backlinks, local, maps, ecommerce, hreflang and more
- 19 specialist agents, with up to 17 running in parallel during a full audit
- 9 optional extensions for external data and image generation
- Python scripts for the Google APIs, page rendering and URL safety checks

How to Install Claude SEO
There are three ways to install it. The plugin route is the one the project now recommends.
Prerequisites
- Python 3.10 or higher. Python 3.10 reaches end of life in October 2026, so use 3.11 or newer for a new install.
- Claude Code installed and working
- Git
- Optional: Playwright Chromium, which the installer offers to set up and which is only needed for rendering single page apps and taking screenshots
- Optional: Google API credentials for Search Console and PageSpeed data
Method 1: Plugin Marketplace (Recommended)
In Claude Code 1.0.33 or later, run these three commands:
/plugin marketplace add AgriciDaniel/claude-seo
/plugin install claude-seo@agricidaniel-claude-seo
/seo setup
In current versions of Claude Code, /plugin install first opens the plugin's details, where you review what it adds and choose a scope before it installs. The plugin install does not run any package manager. /seo setup is the explicit one time step that creates an isolated Python environment and Chromium inside Claude's plugin data folder. Run /seo doctor any time for a read only check of that runtime. This setup step is new since v2 and is the most common thing old tutorials leave out. Auto update is off by default for third party marketplaces, so as the release notes show, update with claude plugin update claude-seo@agricidaniel-claude-seo. If setup fails, the installation guide says to rerun the installer, and that a failed Chromium step only degrades the result: raw fetch analysis still works.

Method 2: Manual Install on macOS and Linux
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
bash claude-seo/install.sh
This copies the main skill to ~/.claude/skills/seo/, the subskills to ~/.claude/skills/seo-*/ and the agents to ~/.claude/agents/seo-*.md, and builds an isolated environment at ~/.claude/skills/seo/.venv/. The README also shows a review then run version of the curl one liner, which is what I would use: download the script, read it, then run it.
Method 3: Windows PowerShell
git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git
powershell -ExecutionPolicy Bypass -File claude-seo\install.ps1
The project dropped the old irm | iex one liner because Claude Code's own guardrails flag piped remote scripts as a supply chain risk. Cloning first lets you read install.ps1 before it runs.
Verify and Uninstall
Open Claude Code and type /seo. If you see the command menu, it works. To remove a manual install, clone the repo and run bash claude-seo/uninstall.sh.
All Claude SEO Commands Explained
The README says there are 34 user invocable commands and shows a table of the core ones, which is the screenshot below. The full reference, with every command and subcommand, is in the repository's docs/COMMANDS.md.

Here is what the main ones do, as the project documents them.
/seo audit (Full Site Audit)
The flagship command. According to the project's command reference, /seo audit https://yoursite.com will:
- Crawl up to 500 pages
- Detect your business type: SaaS, local service, ecommerce, publisher or agency
- Hand the work to up to 17 specialist agents in parallel, nine that always run and eight that run when relevant
- Calculate a health score from 0 to 100
- Write
FULL-AUDIT-REPORT.mdandACTION-PLAN.md

The score weights are unchanged from the version I first covered, which I checked against the current SKILL.md:
- Content Quality: 23%
- Technical SEO: 22%
- On page SEO: 20%
- Schema and Structured Data: 10%
- Performance (Core Web Vitals): 10%
- AI Search Readiness: 10%
- Images: 5%
One cost change matters if you run audits often. Since v2.3.1, five of the judgment heavy agents (content, geo, sxo, cluster and drift) run on Opus and the rest run on Sonnet. The README says plainly that a full audit therefore costs more than it did on v2.3.0, and I have not measured how much more, and you can change any agent's model by editing the model: line in its file under agents/.
/seo page and /seo technical
/seo page is a deep dive on one URL: on page elements, content quality, technical tags, schema, images and Core Web Vitals risks. /seo technical covers nine categories: crawlability, indexability, security, URL structure, mobile, Core Web Vitals (LCP under 2.5 seconds, INP under 200 milliseconds, CLS under 0.1), structured data, JavaScript rendering and IndexNow.
/seo content and /seo content-brief
/seo content scores a page against Google's E-E-A-T framework, using the Search Quality Rater Guidelines as updated in September 2025, and checks how citable it is for AI answers. The README says the scorecard weights Trust highest, following Google's own statement that trust is the most important part. /seo content-brief is newer: give it a topic and it writes a brief with target keywords, an outline and internal link suggestions.
/seo schema
Detects JSON-LD, Microdata and RDFa, validates it against Google's current requirements and generates ready to paste markup. The deprecation list has moved on since my first version. The README says Google stopped showing FAQ rich results for all sites on May 7, 2026, so FAQPage markup no longer earns a rich result, and HowTo rich results were removed in September 2023. If your old schema advice still says to add FAQ markup for the rich result, it is out of date.
/seo geo and the new /seo agentic
/seo geo analyzes how your content performs in AI search: Google AI Overviews, ChatGPT search, Perplexity and Bing Copilot. It scores passage citability, question based heading structure, attribution density and entity presence. The README follows Google's AI Optimization Guide, which says optimizing for generative AI search is still SEO, and it treats llms.txt as not currently a citation lever.
/seo agentic arrived in v2.4.0 in September. It scores a page the way Lighthouse's Agentic Browsing category counts it and checks AI agent access, llms.txt, Markdown delivery, ai-catalog.json and WebMCP. I ran the script behind it against this site, and the output is in the real run section below.
The rest of the toolbox
/seo sitemap: validate a sitemap, or generate one with industry templates and quality gates/seo images: alt text, file sizes, formats, lazy loading and layout shift/seo plan: a phased roadmap for SaaS, local, ecommerce, publisher or agency sites/seo localand/seo maps: Google Business Profile, citations, reviews and map pack analysis/seo programmatic: analysis for sites that generate pages from databases, with safety gates against thin and doorway pages/seo hreflang: validate international tags/seo competitor-pages: "X vs Y" and "alternatives to X" page structures/seo backlinks: link profile analysis from free sources such as Moz, Bing and Common Crawl/seo cluster,/seo sxo,/seo ecommerce,/seo driftand/seo flow: newer commands for keyword clustering, search experience optimization, store analysis, change monitoring with SQLite snapshots, and the project's FLOW prompt set
The Extension System
The core commands work with no API keys. Extensions add live data from outside services, and each one ships its own installer under extensions/. The repository currently has nine: Ahrefs, Banana, Bing Webmaster, DataForSEO, Firecrawl, Matomo, Profound, SE Ranking and Unlighthouse. The README highlights these:
- Banana generates SEO images such as OG previews and blog heroes through Google's Gemini models.
- SE Ranking adds AI share of voice across ChatGPT, Gemini, Perplexity, AI Overviews and AI Mode.
- Profound tracks LLM citations over time.
- Bing Webmaster combines Bing Webmaster Tools with IndexNow submission.
- Unlighthouse runs Lighthouse across many pages locally.
Some of these wrap paid services where you bring your own account. I have not run any of the extensions for this refresh, so I am passing along what the project documents, not what I measured.
seo-cockpit: the newest addition
Version 2.4.2, released on October 4, adds seo-cockpit, an optional companion plugin that needs Claude Code 2.1.287 or later. According to the release notes it includes a spend guard that holds paid SEO API calls against a budget you set, a live audit band, an economy mode and a pane for Search Console, rankings and Core Web Vitals data. If you connect paid data sources, the spend guard is the part I would look at first. I have not tried it.
Setting Up Google API Integration
The Google integration is the most involved setup and gives the richest data. It works in four tiers, so you can stop at whichever one you need:

Tier 0, API key only. Create an API key in the Google Cloud Console, enable the PageSpeed Insights and Chrome UX Report APIs, and you get PageSpeed, CrUX field data and the 25 week CrUX history. No OAuth needed.
Tier 1, OAuth or a service account. Adds Search Console queries, URL Inspection, sitemap status and the Indexing API for eligible pages.
Tier 2, GA4. Adds organic traffic, top landing pages and device and country breakdowns once you configure a GA4 property.
Tier 3, Google Ads token. Adds Keyword Planner volume and competition data.
Run /seo google setup for the credential wizard. The README says credentials live under ~/.config/claude-seo/ with 0600 permissions, and that PDF reports are built with WeasyPrint and matplotlib. The command reference lists /seo google report full for a full PDF and HTML report, and the README describes the PDF of a full site audit as roughly 32 A4 pages, with a cover, an executive summary and recommendations. The security notes in the changelog are worth a read too: the project has patched SSRF bypasses, moved API keys into request headers and added a secret scanning check to its CI.
How the Full Audit Works
When you run /seo audit, the documented sequence is:
- Homepage analysis. The orchestrator fetches your homepage and looks for business type signals: pricing pages suggest SaaS, addresses and phone numbers suggest a local business, product pages and carts suggest ecommerce.
- Business type detection. It picks one of five types, which decides which conditional agents run.
- Specialist agents in parallel. Technical, content, schema, sitemap, performance and visual agents always run. Local, maps, Google and others join when the site calls for them.
- Independent analysis. Each agent works in its own context window with its own tools.
- Scoring. The orchestrator merges the results into the weighted health score.
- Reports. It writes the full report and the prioritized action plan. Each recommendation carries the observation behind it, how it depends on other fixes, a check for how you would know it failed, and a leading indicator.
The README claims an audit takes 10 to 15 minutes, which is the author's figure, and the time depends on site size and which agents run.
Real Audit Results
My own audit of jahanzaib.ai
On July 20, 2026 I ran a full audit of this site with the version of Claude SEO installed on my machine at the time, v1.7.0: eight parallel specialists across 302 sitemap URLs, plus live spot checks. The overall score was 78 out of 100 as a baseline before any fixes. Here is the top of the report it wrote:

By category the scores were Technical SEO 88, Content Quality 74, On-Page SEO 78, Schema 87, Performance 62, AI Search Readiness 73 and Images 80. The report flagged 47 findings, and I fixed about 25 of them in the same session. The performance score came from a hero section that only rendered after JavaScript hydration, which delayed the largest contentful paint. One of the more embarrassing finds was a bug in my own image pipeline that served three oversized images of about 2 MB each, and they were on the earlier version of this very post. The report also noted that scoring was stricter than my April run, so I should not read the changes between audits as a precise trend. The next audit I run will use v2.4.2.
The agent readiness check, run today
For this refresh I ran the v2.4.2 agent readiness script against the live site. It rates checks by priority and prints pass, fail or info for each:

scripts/agentic_check.py against this site on October 7, 2026, rendered as a screenshot.The three top priority checks passed: the content is present without JavaScript, robots.txt is reachable, and it has deliberate groups for AI user agents. One check failed. The script says my llms.txt does not follow the Lighthouse rules because it contains no Markdown links. The README itself says Google does not currently use llms.txt as a citation lever, so I am weighing the fix against the effort and have not changed anything yet.
The same scripts handle the basics too. This second screenshot shows sitemap discovery, the HTML parser and the content scorer running on my Pydantic AI tutorial:

I could not run the slash commands themselves for this refresh, because the nested Claude Code session I used for testing could not authenticate. What you see above is the output of the plugin's own Python scripts, which the audit agents call, not a full agent run.
The project's own showcase
The project website shows three audits of other sites: an ecommerce site at 64 out of 100, a SaaS tool at 58 and a local electrician at 62. Those are the author's results and I have not reproduced them.

How Claude SEO Compares to Paid Tools
The comparison, with prices from the vendors' own pricing pages on October 7, 2026, and checked again on October 11:
| Feature | Claude SEO | Ahrefs | Semrush |
|---|---|---|---|
| Price | $0, plus your Claude Code subscription | Lite $129/mo, Standard $249/mo, a $29/mo Starter plan | SEO plan $139/mo monthly, or $117.33/mo billed annually |
| Site audit | Yes, local, up to 500 pages | Site Audit included | Site Audit included |
| AI search | Checks your own pages for AI readiness | Brand Radar tracks AI prompts, 5 on Lite | AI visibility and sentiment tracking in the plans |
| Backlink and keyword data | Free sources for backlinks, extensions for paid data | Its own web index | Its own dataset |
| Rank tracking | No tracker of its own: Search Console average position, plus one off SERP and Maps checks through paid extensions | Built in, 750 keywords weekly on Lite | Built in, 500 keywords daily on the SEO plan |
| Browser dashboard | No, markdown and PDF reports, plus the optional cockpit pane | Yes | Yes |
| Where your data goes | Stays local by default | Vendor cloud | Vendor cloud |
Both Ahrefs and Semrush now offer MCP access, which matters because Claude SEO can pull from them. Ahrefs lists API and MCP access on its Lite plan and above, while Semrush lists MCP access on its $199 a month Starter plan and above, not on the $139 SEO plan in the table.

Claude SEO wins on cost, schema generation, agent readiness checks and repeatable audits. Paid tools win on proprietary backlink and keyword indexes, a shared dashboard for people who will never open a terminal, and long running rank history. The README makes the same call about Screaming Frog: it crawls deeper and faster at the link graph level, and Claude SEO does not try to replace it.
Other SEO Tools That Connect to Claude
If you came here looking for rank tracking, here is the plain answer first: Claude SEO does not track rankings. Nothing in it checks a list of keywords on a schedule, and the only position data it keeps is the dated file each Maps grid scan saves for later comparison. I checked the README and the source for this on October 10, 2026. The closest it gets:
/seo google gscpulls clicks, impressions, CTR and average position from Search Console for the last 28 days, once you have set up Google credentials.- The
seo-cockpitRankings tab charts that same Search Console average position over time. The cockpit's own code says its rankings come from Search Console. /seo maps gridruns a one off geo grid scan of Google Maps results around a location, and it spends DataForSEO credits to do it. The project calls this geo grid rank tracking, but each scan is a single snapshot you start yourself.- The DataForSEO and SE Ranking extensions can fetch the live results page for a keyword, which is a snapshot, not a history.
Search Console's average position is Google's own number for your site. Google defines it as the topmost position of your page, averaged across the searches where it appeared, so it may not match a manual check of one keyword from one place.
If you want positions for a fixed set of keywords tracked over time, the tracking has to run in a service with its own schedule, and Claude reads the results through that vendor's MCP server. These are the tools I could verify on the vendors' own pages or official repositories. Prices were checked on October 10, 2026 and again on October 11, and the table describes what each vendor documents, not my own measurements.
| Tool | What it is | What it covers | Cost (checked October 11, 2026) | How it connects to Claude |
|---|---|---|---|---|
| Ahrefs MCP | Official hosted MCP server | Site Explorer, Keywords Explorer, Rank Tracker, Site Audit, Brand Radar and your Search Console data, grouped by tool | Lite plan and up, from $129/mo; each call uses API units from your plan | Claude's connector directory on web and desktop, or claude mcp add ahrefs https://api.ahrefs.com/mcp/mcp -t http in Claude Code |
| Semrush MCP | Official hosted MCP server, listed as a Semrush app in Claude | SEO API data on keywords, domains and backlinks, Trends API data with a Trends subscription, and read only access to your projects, which include Position Tracking and Site Audit | The pricing page lists MCP access from the Starter plan, $199/mo billed monthly; eligible plans include 50,000 API units | Claude's connector directory, or claude mcp add semrush https://mcp.semrush.com/v2/mcp -t http in Claude Code |
| SE Ranking MCP | Official hosted MCP server | Keyword, backlink, domain and SERP data, your projects with rank tracking and audits, local rankings, and AI visibility through SE Visible | Core plan $129/mo billed monthly, with API access and 2,000 keywords tracked daily; a standalone pay as you go API from $50 for 250K credits; 14 day free trial | Claude's connector directory, or claude mcp add --transport http se-ranking https://api.seranking.com/mcp in Claude Code |
| DataForSEO MCP | Official open source server (Apache 2.0) with a hosted endpoint | Raw SERP results, keyword volume and difficulty, backlinks and on page checks, with no dashboard | Pay as you go with a $50 minimum payment; one Google organic results page starts at $0.0006 in the standard queue and $0.002 live | Claude Desktop config through npx, or a remote or local server in Claude Code |
| Moz MCP | Official hosted MCP server, in beta for account owners | Moz Data: site metrics, links, keywords and ranking data, plus a separate Moz Local endpoint for listings | Moz Data needs a Moz Data API subscription, from $20/mo for 3,000 rows; Moz Local needs its own Moz Local subscription | Moz Data and Moz Local in Claude's connector directory, or claude mcp add with Moz's URL in Claude Code |
| mcp-gsc for Search Console | Community project, MIT licensed, about 1,900 GitHub stars as of October 10, 2026. I found no Search Console MCP server published by Google | Queries and pages with clicks, impressions, CTR and average position, URL inspection, and listing, submitting and deleting sitemaps | Free; you create your own Google Cloud credentials | Claude Desktop config, plus other MCP clients |
| Screaming Frog SEO Spider MCP | Official, built into SEO Spider version 24, paid license only | Runs crawls on your own machine, then analyzes and exports the crawl data | £199 per license per year | Claude Desktop through Screaming Frog's MCP extension, in STDIO or Streamable HTTP mode |
Matching the tool to the job
- Rank tracking with history. Ahrefs, Semrush and SE Ranking run the tracker on their side, on their own schedule, and Claude reads the results when you ask. On the plans in the table that means 750 keywords updated weekly on Ahrefs Lite, 500 keywords tracked daily on Semrush Starter and 2,000 keywords tracked daily on SE Ranking Core. Ahrefs Lite and SE Ranking Core cost the same $129 a month billed monthly, so for a fixed keyword list read through Claude, SE Ranking tracks the most keywords of the three and updates them daily, where Ahrefs updates weekly.
- Keyword and SERP data without a seat licence. DataForSEO bills per request and the Moz API by a monthly row allowance, and both cost less to start than the suites. DataForSEO gives you a live rank check for any keyword, but keeping the history is up to you.
- Free. Search Console, read through a community server like mcp-gsc or through Claude SEO's own
/seo google gsc, gives you an average position for each query and day on your own site at no cost. It is still an average, it only counts searches where your page got an impression, and Google leaves rare queries out for privacy, as it explained in a 2022 deep dive on the data. So it is not the same as a fixed daily rank check. - Deep technical crawls. Screaming Frog, run from Claude Desktop, with Claude SEO doing the scoring, schema and AI search checks around it.
- Audits from Claude Code with no subscription. Claude SEO itself, with outside data added only when you need it. Its extensions already cover Ahrefs, DataForSEO and SE Ranking.
Version History and Project Activity
The old version of this guide covered v1.0.0 to v1.8.0 in 54 days. In the six months since, the project shipped 19 more releases. The ones worth knowing:
- v2.0.0 (May 2026): a security and AI search major release, with a shared headless renderer for single page apps and an SSRF hardening pass
- v2.1.0 (May): a currency refresh for the May 2026 core update and the retirement of FAQ rich results
- v2.2.x (June to September): credential and path traversal fixes, a managed cross platform runtime, Lighthouse guidance including the Agentic Browsing category
- v2.3.1 (September 10): five judgment heavy agents moved to Opus
- v2.4.0 (September 23): the new
/seo agenticcommand - v2.4.1 (September 29): a Google currency patch for the September 2026 spam update
- v2.4.2 (October 4): seo-cockpit and Google sign in
The GitHub API shows 18,470 stars, 2,702 forks, 28 contributors and 29 release tags as of October 7. When I checked the releases again on October 11, v2.4.2 was still the newest, and the last commit on the main branch was dated October 4. That is a very different project from the one I described in April, when it had about 3,000 stars. If you installed v1 and never updated, you are missing most of the above, and the project has a v1 to v2 migration guide.
When to Use Claude SEO (and When Not To)
Use it when:
- You want a thorough audit without a monthly SEO subscription
- You want checks for AI search and agent readiness that most older tools still lack
- You prefer terminal workflows and markdown reports you can keep in git
- You want schema generated and validated in one step
- You already use Claude Code and want repeatable audits
Skip it, or pair it with a paid tool, when:
- You need proprietary backlink or keyword databases
- Your team needs a shared browser dashboard for non technical stakeholders
- You need months of rank history, since this is an audit tool, not a monitor
- You run pages that depend on heavy client side hydration timing, which the README lists as a known source of noisy findings
- You are not comfortable working in a terminal
Getting Started: Your First Audit
- Install Claude Code if you don't have it.
- In Claude Code, run
/plugin marketplace add AgriciDaniel/claude-seo, then/plugin install claude-seo@agricidaniel-claude-seo. - Run
/seo setuponce, then/seo doctorto confirm the runtime is ready. - Run
/seo audit https://yoursite.com. - Read
ACTION-PLAN.mdfirst and fix the critical items before the rest. - Check a fixed page with
/seo page https://yoursite.com/fixed-page. - Add the free Google Tier 0 API key for real PageSpeed and CrUX data.
Frequently Asked Questions
Is Claude SEO really free?
Yes. The core plugin is MIT licensed and open source. You need a Claude Code subscription, which you pay Anthropic separately, and a full audit uses more of it than it used to because several agents now run on Opus. Optional extensions that wrap paid services need their own accounts and carry their own pricing.
Does my data get sent to external servers?
Not by default. The README says the core plugin makes no third party API calls and has no telemetry, though audits still fetch the URLs you give them. Data only leaves your machine when you turn on an extension or the Google API integration, or when Claude Code itself sends your prompts to Anthropic.
How is it different from Screaming Frog?
Screaming Frog is a dedicated crawler that goes deeper and faster across big link graphs. Claude SEO adds content quality scoring, schema generation, AI search and agent readiness checks, and local SEO. The README says it does not try to replace a crawler, so the two complement each other.
Does it work on React, Next.js and other single page apps?
Yes, with limits. Since v2 it uses a shared headless renderer built on Playwright that detects empty app shells and switches to a rendered fetch. The README warns that pages whose content depends on scrolling or user interaction can still produce noisy findings.
Which version should I install, and how do I update?
Install the current public release from the AgriciDaniel/claude-seo marketplace, which is v2.4.2 at the time of writing. Updates are not automatic for third party marketplaces, so run the plugin update command periodically. If you are still on v1, read the migration guide because v2 changed defaults.
What is the difference between /seo audit and /seo page?
The audit crawls up to 500 pages and runs specialist agents in parallel, then scores the whole site. The page command goes deep on one URL. Use the audit for a site wide baseline and the page command to confirm a fix.
Can I use it on client sites?
Yes. The reports are plain markdown, and the Google report command can produce PDF and HTML with charts. Only run the agent user agent matrix on sites you own or have permission to test, as the project's agentic skill notes.
For related reading, see my guide to when to use AI agents versus plain automation, the walkthrough of building production agents with LangGraph, and the free AI crawler check for a quick look at which AI bots can reach your site. If you want a second opinion on your own audit results, get in touch.
Related reading
- How to track AI search visibility: measure whether the fixes show up in AI answers.
- Answer engine optimization service: how I handle AEO and SEO for clients.
- AEO glossary: 32 answer engine terms in plain words.
- Claude Code cheat sheet: the commands and settings you use with this tool.
- How to add MCP servers to Claude Code: connect more tools to the same session.
Last updated
November 1, 2026
Category
Implementation
Jahanzaib Ahmed
AI Systems Engineer & Founder
AI Systems Engineer with 126 production systems shipped. I run AgenticMode AI (AI agents, RAG systems, voice AI) and ECOM PANDA (ecommerce agency). I build AI that works in the real world for businesses across home services, healthcare, ecommerce, SaaS, and real estate.
Want this built for your business? See Managed AEO and SEO, or browse the free tools and guides.
Related articles

How to Get Claude Code at the Lowest Cost: Plans and Free Routes
Claude Code installs for free but needs a paid Claude plan or API credits to run. Every plan's price, how the usage limits work, and the few free routes that are real.
How toImplementationClaude Code
How to Remove an MCP Server From Claude Code When It Keeps Coming Back
Claude Code lists the same MCP server in up to five places, so one remove command often leaves it running. Here is how to find every copy, remove it by scope and confirm it is gone.
How toImplementationClaude Code
How to Track AI Search Visibility: What Free Reports Show and What Seven Paid Tools Cost
What the free Bing and Search Console AI reports show today, how to set up a free routine, and what seven AI search visibility tracking tools cost as of October 17, 2026.
How toImplementationAI Visibility