Ryze-AI-Adgent/open-seo-mcp-skills: SEO skills for Claude that run on your own GSC and GA4 data

The problem: SEO tools that are free code around paid data

The README makes an argument rather than a feature list. Most open-source SEO tooling, it says, is a UI over the DataForSEO API: you bring your own key and pay per request, or you pay a hosted subscription plus what the README describes as a roughly 28% markup on every data call. The code is free; the data is not.

This repository takes the opposite position. Rankings come from your actual Google Search Console, so positions and clicks are the ones Google reports rather than estimates derived from scraping SERPs. Traffic comes from GA4, including referral traffic from ChatGPT, Perplexity, Claude and Gemini. Keyword volumes come from Google Ads keyword planner data. Competitor keywords, backlinks and SERPs come from DataForSEO, already wired into the Ryze connector, so there is no key to manage. Ahrefs and Semrush are described as connectable if you already pay for them.

The audience is narrow on purpose: someone working inside Claude Code who has access to their own Search Console and GA4 properties and wants answers phrased as questions rather than as a dashboard to click through. If you do not use Claude, most of this repository is inert.

How the skills, the MCP connector and your Google accounts fit together

There are two separable pieces. The first is the Ryze MCP server at https://connector.get-ryze.ai/mcp, an HTTP transport endpoint. You sign in, pick a workspace, and connect your Google Search Console, GA4 and ads accounts once. That server holds the credentials and exposes tools to Claude.

The second piece is the skills themselves, which live under the skills/ directory and are installed into Claude as a plugin. A skill is a written procedure: it tells Claude which MCP tools to call, in what order, and how to shape the answer. The README lists eight of them, each mapped to a data source. seo-audit covers indexation, CTR anomalies, decaying pages and quick wins from GSC plus GA4. keyword-research goes from seed to clustered plan using Google Ads and DataForSEO. rank-tracking compares position movers between two periods using GSC alone. competitor-gap finds keywords a competitor ranks for that you do not. backlink-check profiles referring domains against a competitor. ai-visibility reports which AI engines send traffic and which pages they cite, from GA4 AI-referral reports. content-brief builds a SERP-driven brief with headings, entities, questions and internal links. seo-vs-ads finds queries you pay for that you already rank for organically, joining GSC against Google Ads.

For anyone writing their own skill, the README documents the naming convention: Ryze MCP tools are namespaced as provider__tool, with examples such as google_search_console__runRawSearchAnalytics and google_ads__generateKeywordIdeas. When a request shape is unclear, the native__get_provider_docs tool takes a provider name such as dataforseo or google_search_console and returns the official API documentation. The README warns that tool availability can vary by workspace and says every skill in the repository states its fallback.

Installing the plugin and running a first audit

The README gives a three-command quick start for Claude Code. The first adds the MCP server over HTTP transport:

claude mcp add ryze --transport http https://connector.get-ryze.ai/mcp

After this, the connector should appear in Claude's MCP server list. The README notes that the plugin bundles the connector, so this step is optional in Claude Code.

The second and third commands register the plugin marketplace and install the skills:

claude plugin marketplace add Ryze-AI-Adgent/open-seo-mcp-skills
claude plugin install open-seo-mcp-skills@ryze

There is an alternative if you would rather not use the plugin system. The README shows cloning the repository and copying the skill directories straight into your Claude skills folder:

git clone https://github.com/Ryze-AI-Adgent/open-seo-mcp-skills && cp -r open-seo-mcp-skills/skills/* ~/.claude/skills/

If you are in the Claude app rather than Claude Code, the README describes a manual path: Settings, then Connectors, then Customize, then add a custom connector with the name Ryze AI and the URL https://connector.get-ryze.ai/mcp. You then sign in, pick a workspace, and connect your Google accounts once. The README points to a full guide at https://www.get-ryze.ai/how-to-connect-claude-to-google-meta-ads-mcp.

Once installed, the first real use is a prompt rather than a command. The README suggests asking for an SEO audit on your site, and gives two other example phrasings: what keywords a competitor ranks for that you do not, and how much you are paying for clicks you would get free. Expect the audit to come back as a written analysis with indexation issues, CTR anomalies and decaying pages, since those are the four things the seo-audit skill is documented to cover.

The dependency you are actually taking on

The skills are MIT and the repository says they are free forever. The data access is not in this repository at all. It runs through your own Ryze workspace and your connected accounts, which means the connector at connector.get-ryze.ai is a required piece of infrastructure sitting between Claude and your Google properties.

That has a few consequences worth naming. The README does not document a self-hosted deployment of the MCP server, so there is no described path to running the data layer yourself. If the connector is unreachable, the skills have nothing to call. The README also does not document rate limits, uptime commitments, or what happens to stored OAuth tokens when you disconnect a workspace. Those are questions to put to the provider, not gaps in the repository you can read your way out of.

There is also a scope limit inherent in the design. Because rankings and traffic come from your own GSC and GA4, the tool can only report on properties you have verified access to. It will not tell you where a site you do not own ranks, beyond whatever DataForSEO returns for that domain. For pure competitive research on sites outside your account, the GSC-backed skills simply do not apply.

A second, quieter limitation: the README states that tool availability can vary by workspace. That means a skill that works for one user may hit a missing tool for another, and the fallback behaviour described in each skill is what you get instead. If you are evaluating this for a team, test on the workspace you actually intend to use.

How it differs from Semrush, Ahrefs and DataForSEO wrappers

The README's own comparison table sets out the difference. Against Semrush and Ahrefs, the split is on where the numbers come from. Those platforms estimate rankings from SERP data and bundle the data cost into a subscription the README puts at $120 or more per month. This project reads your real GSC positions and your real GA4 traffic instead, and the tool itself costs nothing under MIT. It also runs inside Claude, which the comparison table lists as a capability the established platforms do not have in that form.

Against the category the README calls OpenSEO-style wrappers, the difference is where the money goes. Those are free code with a DataForSEO bill attached, or a hosted version with a markup on each call. Here, DataForSEO is already wired into the Ryze connector, so there is no key to manage and no markup layer to build. The README lists Ahrefs and Semrush as connectable if you already pay for them, which is a reasonable hedge: you keep the subscription you have and let the skills read it.

The honest read is that this is not a replacement for a rank tracker on domains you do not own, and it is not a replacement for a backlink index with its own crawl. It is a way to interrogate data you already have, plus a DataForSEO account you are not managing directly, from inside a chat interface.

Maintenance, licensing and what the repository does not promise

The repository is not archived, and the last push was on 2026-09-09, which is recent enough that the code is not stale. There are no retrieved releases, so there is no versioned changelog to read and no tagged upgrade path. Updating means pulling the plugin or re-cloning and re-copying the skills directory. The primary language is Shell, and the repository is small: a plugin manifest directory, a LICENSE, a README, an install.sh, an image, and the skills directory. That is a footprint you can read end to end before trusting it.

The licence is MIT, which covers the skills and the repository. It does not cover the data. The README is explicit that data access runs through your own Ryze workspace and your own connected accounts, so the terms that govern your Search Console, GA4 and ads data are the ones attached to those accounts and to the connector, not the MIT grant. If you fork the skills and ship them, the MIT terms travel with the code; nothing about the connector does. This is a description of what the licence text covers, not legal advice, and the connector's own terms are the place to look for the rest.

The maintenance cost that matters here is not in the repository. Eight skills that call a namespaced tool surface will drift when that surface changes, and the README's note that availability varies by workspace means there is no single environment in which a maintainer can verify every skill at once.

Editorial conclusion

Adopt open-seo-mcp-skills if you already live in Claude Code and want audit, rank-tracking and competitor-gap answers computed from your own GSC and GA4 properties rather than from SERP scrapes. Skip it if you need a standalone dashboard, if you cannot connect Google accounts to a third-party MCP endpoint, or if you want a tool that keeps working when that endpoint does not. Before committing, install the plugin, run one audit against a property you know well, and check the answer against the same numbers in Search Console.

Frequently asked questions

Official sources

1. Issues
2. License: MIT
3. Project website
4. README
5. Ryze-AI-Adgent/open-seo-mcp-skills on GitHub