analytics-mcp

analytics-mcp is a Model Context Protocol server that puts GA4, Cloudflare Web Analytics, Vercel Analytics and Google Search Console behind one normalized tool surface. It is built to teach the model what the numbers mean, with each metric carrying a plain-language meaning and the server knowing which discrepancies between trackers are structural. It can be deployed remotely using StreamableHTTP and OAuth 2.1 on Vercel, or run locally over stdio for Claude Code, Cursor, and Claude Desktop. The server exposes read-only tools such as list_sites, list_sources, get_schema, query, query_raw, explain_discrepancy, validate_query, seo_opportunities, explain_ctr_gap, ai_referrals, inspect_page, and page_changes. Canonical metrics include pageviews, sessions, visitors, clicks, impressions, ctr, and position, each mapped to native names per source. Search analysis computes a click-through curve from the user's own Search Console rows, and AI visibility reports sessions arriving from AI assistants by engine while stating that arrivals are not citations. inspect_page fetches a page and reports mechanical findings such as title length, missing description, duplicate headings, canonical, noindex, and non-200 status. page_changes records page state only when a content hash moves, and the store is optional. Setup requires Node 20+ and pnpm, with SITES_CONFIG describing sites and credentials configured per source.