open-seo-mcp-skills: A Free Claude Code SEO and GEO Toolkit Built on Your Own Data

Seven distinct skills ship in one repository: seo-audit, keyword-research, rank-tracking, competitor-gap, backlink-check, ai-visibility, and content-brief plus seo-vs-ads. That's the actual shape of open-seo-mcp-skills, an open source Claude skill pack built on the Ryze MCP connector, and it's grown to 423 GitHub stars, up 78 in the past week, the kind of trajectory this site's own scan has watched building for several days before it cleared the bar for a full writeup.

The pitch is straightforward: instead of paying for a dashboard that estimates your rankings and traffic from a third-party crawl, connect Claude directly to the accounts that already hold your real numbers, Google Search Console, GA4, Google Ads, and DataForSEO, and let it run the analysis on data you actually own.

What the seven skills actually cover

seo-audit looks at your own site's indexation status, flags pages with unusual click-through-rate patterns, and surfaces pages that are declining in a way that's worth investigating. keyword-research builds a seed-to-plan keyword strategy pulling real search volume and CPC data rather than a generic estimate. rank-tracking compares your Search Console positions across two time windows and reports what moved. competitor-gap finds keywords a competitor ranks for that you don't. backlink-check profiles a domain's backlink portfolio and referring-domain mix through DataForSEO. content-brief pulls a SERP analysis for a target keyword into a brief a writer can actually use, and seo-vs-ads shows where paid and organic coverage overlap so ad spend isn't propping up a keyword you already rank for.

The one that stands out against most SEO tooling on the market right now is ai-visibility: it tracks which AI answer engines are sending referral traffic and, where detectable, which of your pages they're citing. That's directly useful against a problem this scan's own citation-gap thread has been circling for weeks, the fact that a citation drop or spike in ChatGPT or Perplexity can look identical to a ranking change in Search Console unless something is specifically watching for it. A study covered here just this week found that citation behavior varies sharply by query type, evaluation-stage B2B searches favor product pages while broad discovery queries favor Reddit and YouTube, which is exactly the kind of split a tool like this can help a site figure out for its own traffic rather than trusting an industry-wide average.

How it actually works

Getting this running takes two steps: connect the Ryze MCP connector inside Claude's settings, then install the skill pack either through the Claude plugin marketplace or by copying the skill files into place manually. Once connected, each skill invokes namespaced MCP tools, google_search_console__runRawSearchAnalytics is a representative example, that query your linked accounts directly rather than pulling from a cached or estimated dataset somewhere else. Ask Claude to run a rank-tracking check, and it queries Search Console for exactly the date ranges and pages you specify, not a rough approximation refreshed on someone else's schedule.

In practice, a session looks less like running a script and more like a conversation: ask Claude to check which pages lost search visibility this month, and it queries Search Console directly, cross-references the affected URLs against GA4 traffic, and returns a ranked list with a plain-language read on what changed, rather than a dashboard full of charts that still need interpreting. Point it at a target keyword with content-brief and it pulls the current SERP, works out what the ranking pages actually cover, and hands back a brief instead of a spreadsheet of metrics.

That data dependency is also the tool's real limitation. Everything meaningful here requires the underlying accounts to exist and be connected: no Search Console history means no rank tracking, no DataForSEO access means no competitor-gap or backlink data. This isn't a tool that works out of the box for a brand-new site with no search history yet, and the competitor and backlink features specifically route through DataForSEO's paid API, so "free" describes the skill code and license, not necessarily every data source behind it.

License, traction, and who it fits

The repository is MIT-licensed, maintained by Ryze-AI-Adgent, and free to fork or modify. The star count, 423 as of this week and climbing at roughly 20% growth week over week, is a reasonable signal of developer interest, though it says nothing about whether any individual skill's output holds up against a paid alternative like Semrush or Ahrefs in day-to-day use, since GitHub stars measure curiosity, not verified results.

This fits a marketer, founder, or in-house SEO who already has Search Console and GA4 set up and wants Claude doing the analysis work instead of clicking through a separate dashboard every week, especially anyone who cares about the AI-citation side of visibility as much as traditional rankings. It's not the right pick for an organization that wants Semrush or Ahrefs feature parity, a polished client-facing report, or subscription-backed support if something breaks. It's also not useful the day a site launches, before there's enough Search Console history for rank-tracking or seo-audit to have anything real to analyze. For an existing site with real traffic and a marketer willing to connect a few accounts once, it's a genuinely broader toolkit than most single-purpose SEO skills circulating right now, seven distinct jobs in one install instead of one tool per task.