Three SEO tools to connect to Claude Code

Commented SEO? These are the three tools from the reel, with setup and the costs that didn't fit in the video. They help you find opportunities, research a market and review your pages. They are not a one-for-one replacement for Semrush or an SEO specialist.

On 3 October 2026, Semrush's pricing page listed its entry SEO plan at US$139/month billed monthly, or US$117.33/month billed annually. Its help article still calls the tier SEO Toolkit Pro and lists US$139.95/month. These pages disagree on name and monthly price: check checkout before buying. Some tools below also cost money.

Before you start

You need Claude Code, your own access to it, and a terminal. MCP is a connection that lets Claude read outside data; a skill is task-specific instructions and scripts. Install only what you need. Never paste credential files or API keys into chat or GitHub.

1. GSC MCP: find pages already getting seen

What it does: reads queries, clicks, impressions and average positions from your own Google Search Console property. Ask which searches put your pages near page one, then review those pages' titles and content. Average position is not a guarantee of a ranking gain.

Who it's for: a site owner with a verified Search Console property. Without access to that property's data, the connector has nothing to analyse.

Install in Claude Code: follow the repository's OAuth steps: create a Google Cloud project, enable the Search Console API, create a Desktop app OAuth client and download its JSON. Install uv to get uvx. Replace the placeholder with the absolute path to your JSON; keep it private.

claude mcp add gsc -e GSC_OAUTH_CLIENT_SECRETS_FILE=/absolute/path/client_secrets.json -- uvx mcp-search-console

Restart Claude Code, open /mcp and approve Google sign-in when prompted. If connection fails, check which uvx and the repository's setup notes. Example request, not tested here: “List my Search Console properties. For [site], show queries with impressions and average positions from 8 to 20, their landing pages and clicks. Suggest a specific page improvement; do not change anything.”

Cost: mcp-gsc is free MIT-licensed code; you need your own property, Google Cloud setup and Claude Code access. The hosted Advanced GSC service advertised in its README is separate.

2. OpenSEO: research outside your own site

What it does: keyword research, rank tracking, competitor research, backlink overviews (sites linking to you), and site audits. Use it to decide what page to create, compare search competitors and find technical issues. Keyword and backlink data come from a third-party provider.

Who it's for: someone who needs market research beyond first-party Search Console data. OpenSEO is MIT-licensed.

Install in Claude Code: enter the two commands from the official plugin instructions inside Claude Code, one at a time:

/plugin marketplace add every-app/open-seo

/plugin install openseo@openseo

Run /reload-plugins if prompted; open /mcp to authorize OpenSEO. The plugin adds its data connection and SEO skills. Try /openseo:keyword-research after setting up a project. Example request, not tested here: “Research keywords for [product] in [country]. Show search volume, competition and data sources. Group page ideas by intent; don't invent numbers.”

Real cost: the self-hosted app has no software subscription, but needs Docker and a pay-per-use DataForSEO key for keyword/backlink data. Its Docker setup disables app authentication: never expose it publicly without your own protection. The hosted plan starts at US$10/month, including US$10 of usage, with a limited trial credit. Extra data usage and Claude Code may cost more. Not “totally free.”

3. GEO skill: review whether your page is easy to cite

What it does: geo-seo-claude checks content, crawler access, structured data and other signals, returning a score and suggested fixes. GEO is the work of making pages understandable to AI answer engines such as ChatGPT, Claude and Google AI. A score does not verify that an engine cited you or guarantee a citation.

Who it's for: people with a public page who want to improve its answers, evidence and attribution. This is a free MIT-licensed skill, not a market-data subscription.

Install in Claude Code: have Git, Python 3.8+ and Claude Code. Read any installer before running it. The project's manual macOS/Linux installation is:

git clone https://github.com/zubair-trabzada/geo-seo-claude.git

cd geo-seo-claude && ./install.sh

For Windows, use Git Bash with ./install-win.sh as documented by the project. Restart Claude Code and try /geo audit https://yoursite.com on your own public site. Example request, not tested here: “Audit [URL] for GEO. Show evidence for each issue, distinguish fixable issues from hypotheses, and don't promise AI citations.”

Cost: the code is free; Claude Code and optional external services have their own costs. Auriti-Labs/geo-optimizer-skill is another MIT-licensed option; its external citation checks can require paid API keys.

Quick questions

Do I need all three? No. Each answers a different question. Start with your own data before paying for market research.

Will they get me onto page one or cited by AI? No. They surface evidence and suggestions; you still make and measure the changes.

Are they the same as Semrush? No. Their data coverage, support and pricing differ.

Sources and limitations: fact-check notes. Checked 3 October 2026.