Just published @automatelab/seo-performance-mcp on npm. It is a free, MIT-licensed Model Context Protocol server for the part of SEO that starts after the post goes live: triage. Which URLs are decaying, which lost an AI citation, which sit at position 7 with a low CTR and need a title rewrite.
The MCP exposes eight tools. posts.snapshot pulls a 30/60/90-day window per URL across Google Search Console, GA4, Matomo, Microsoft Clarity, and an optional AI-citation source. posts.decay_curve buckets 12 weeks of GSC clicks, impressions, and position into weekly windows and classifies the trend as decay, plateau, or growth. posts.verdict runs a rules-based engine over the snapshot and curve and emits one of six calls per URL (refresh, expand, merge, kill, double_down, hold) with the reason codes that drove the call and a 0-1 confidence score. cohort.report runs the verdict engine across a tag-filtered cohort and returns a ranked table. gsc.quick_wins scans Search Console for (page, query) pairs at positions 5-15 where CTR sits below the position-expected curve.
Read-only by design. Every signal source is optional - start with Search Console alone if that is what you have. The MCP works with Claude Code, Claude Desktop, Cursor, Cline, Continue, and any MCP-compatible host. Three discoverable prompts ship with the server (audit_cohort, find_quick_wins, citation_loss_sweep) so no skill loader is required.
Install:
npx -y @automatelab/seo-performance-mcp
Landing page (full tool reference and env vars): https://automatelab.tech/products/mcp/seo-performance-mcp/
GitHub: https://github.com/AutomateLab-tech/seo-performance-mcp