- Claude is exactly as good as the data you hand it. Map each source to the skill built for its shape: GSC, Screaming Frog, Clarity, Bing and GBP.
- The one integration worth configuring: Screaming Frog v24+ ships an MCP server, so Claude can drive the crawl itself.
- No API keys by design. Export-and-paste is audit-proof, needs zero standing infrastructure, and fails safely when data is missing.
Claude is exactly as good as the data you hand it. Here is every source in a working SEO stack, mapped to the skill that consumes it and what comes back out.
Connecting Claude to your SEO stack means matching each data source to the skill built for its shape: Search Console exports feed the diagnosis skills, Screaming Frog feeds the crawl analyser, Clarity feeds the AI bot auditor, and Bing's AI reports feed citation analysis. No API keys, no middleware. Exports and paste, or live fetches.
The source-to-skill map
| Data source | What to export | Skill that consumes it | What you get back |
|---|---|---|---|
| Google Search Console | Performance, any range | GSC Quick Start | Highest-value pages to fix first |
| Queries with page dimension | Cannibalization Detector | Per-query verdicts where your own pages compete | |
| Two page exports, same length, a year apart | Content Decay Detector | Which pages are bleeding, why, and the right treatment each | |
| Compare view around an update window | Winner or Loser | Per-page update verdicts with the intent-fit diagnosis | |
| Screaming Frog | Crawl export, or MCP mode on v24+ | Screaming Frog Analyser | Prioritised fix plan weighted by link equity and indexation |
| Microsoft Clarity | AI Bot Activity dashboard, 30 days | Clarity AI Bot Auditor | Per-bot verdicts: healthy, blocked, throttled, over-scraping |
| Bing Webmaster Tools | AI Performance report | Citation Share Auditor | Your slice of AI answers and the move that grows it |
| Google Business Profile | GBP collection in GA | GBP Analytics Connector | Monthly local actions report with trends and flags |
| Nothing but a URL | None | Technical SEO Audit, Website Score, AI Visibility | Audits and baselines from live fetches |
| Your website's git repo | None, run it in the repo | Claude Code SEO | SEO_REPORT.md plus fixes staged as commits |
The one genuine integration: Screaming Frog MCP
Most of this stack is export-and-paste by design. The exception worth setting up: Screaming Frog v24+ ships an MCP server, which means Claude can drive the crawl itself, pull the data and analyse it in one conversation. The Screaming Frog Analyser skill detects which mode it is in and runs the same framework either way. If you crawl weekly, the setup pays for itself the first Monday.
Why export-and-paste beats API middleware
Three reasons this stack has no API keys in it. First, exports are audit-proof: the exact rows Claude analysed sit in the conversation, so any finding can be checked. Second, zero standing infrastructure: nothing to break, rotate or pay for. Third, access control stays where it belongs, in each tool's own permissions instead of a token with god-mode scopes.
The pattern also fails safely. A skill that needs the page dimension and does not get it stops and tells you exactly how to re-export, because guessing at cannibalisation without page data would be worse than not answering.
A weekly routine that uses the whole stack
- Monday, 15 minutes: GSC Quick Start on last week's export. It names the pages worth your attention.
- After every content push: Information Gain on the draft before it ships, Website Score on the page after.
- Monthly: Clarity AI Bot Auditor and Citation Share Auditor together. One tells you whether AI systems can reach you, the other whether they cite you.
- Every core update: Winner or Loser on the compare export, before anyone panics in Slack.
Start with whichever source you already have open in a tab. The setup guide covers the install, and the directory has every skill named above.
FAQ
Can Claude read Google Search Console directly?
Not natively, and it does not need to. GSC exports paste straight in: the Performance report with the page dimension added covers most jobs. Four skills in the library are built specifically around GSC export shapes, from quick-start triage to core update forensics.
Does Claude work with Screaming Frog?
Two ways. Screaming Frog v24 and later ships an MCP server, so Claude can drive the crawl itself when the connection is configured. Without it, export the crawl and paste: the Screaming Frog Analyser skill handles both modes with the same analysis framework.
What about Ahrefs and Semrush data?
Export and paste works today: keyword lists feed the clustering and intent skills, backlink exports feed competitive analysis. There is no dedicated skill per paid tool yet, deliberately, because the free data sources cover the highest-value workflows first.
Do I need all these tools to start?
No. Half the library runs on nothing but a URL, fetching what it needs live. Start there, then add GSC exports the day you want cannibalisation or decay analysis. The stack grows one data source at a time, each one switching on new skills.
Open whichever tool you already have in a tab, export per the map above, and hand it to the matching skill from the directory. Monday's 15-minute GSC triage is the easiest place to start.