Half the surface is raw data and half is analyses that would otherwise be a pivot table and an afternoon — cannibalization, striking distance, content decay. That is what makes "which pages should I refresh?" a question you can actually ask. Two notes worth carrying in. Search Console data lags by two days, which the tools account for. And the project is candid about a risk inherent to any tool that surfaces third-party text: spam queries in your Search Console arrive as content the model reads, so treat the data as untrusted input, not instructions.
A local MCP server that connects an assistant to your Semust account, which in turn holds your connected Search Console, Analytics, Ads, Yandex Metrica and Bing Webmaster data. The server runs on your machine and calls the Semust API directly, so the API key never travels to the model — only the resulting data does.
- 56 tools, all read-only, starting with the project list that tells you which integrations are actually connected — `list_projects`
- Search Console data as data: top keywords and pages with growth metrics, and a daily traffic overview with before-and-after comparison — `get_keywords`, `get_pages`, `get_performance`
- Search Console as finished analyses rather than raw rows: keyword cannibalization, a monthly report with year-on-year comparison, long-tail opportunities, winners and losers, question queries, content decay, page-two striking distance, low-CTR items and thin content — `report_cannibalization`, `report_monthly_summary`, `report_long_tail`, `report_winner_loser`, `report_questions`, `report_content_decay`, `report_striking_distance`, `report_low_ctr`, `report_thin_content`
- Google Analytics across the usual cuts — overview metrics with comparison, traffic sources by channel and campaign, top and entry and exit pages, geography, devices and key events — plus trend, landing-page conversion, channel comparison and engagement-tier reports — `ga_get_overview`, `ga_get_traffic_sources`, `ga_get_pages`, `ga_get_geo`, `ga_get_devices`, `ga_get_key_events`, `ga_report_traffic_trends`, `ga_report_landing_page_performance`, `ga_report_channel_comparison`, `ga_report_engagement_analysis`
- Google Ads: campaign performance, daily cost and conversions, geography, devices and the search terms that actually triggered your ads, with spend and conversion-breakdown reports on top — `gads_get_campaigns`, `gads_get_daily_metrics`, `gads_get_geo`, `gads_get_devices`, `gads_get_search_terms`, `gads_report_performance_overview`, `gads_report_spend_analysis`, `gads_report_conversion_breakdown`
- Rank tracking with position history, visibility score, distribution and a gainers-and-losers report — `rank_get_keywords`, `rank_get_overview`, `rank_get_history`, `rank_report_gainers_losers`, `rank_report_distribution`
- AI Overview visibility as its own surface: whether you appear, per-keyword presence and positions, and which domains are being cited instead — `ai_rank_get_overview`, `ai_rank_get_keywords`, `ai_rank_get_domains`
- Indexing health with a score, coverage and warnings, plus the daily history behind it — `indexing_get_status`, `indexing_get_history`
- Yandex Metrica and Bing Webmaster with their own full sets, including Yandex search phrases and Bing crawl issues, link counts and per-URL traffic — `ym_get_overview`, `ym_get_search_phrases`, `bing_get_query_stats`, `bing_get_crawl_issues`, `bing_get_link_counts`, `bing_get_url_traffic`
A Semust account on an active subscription, with at least one project that has Google Search Console connected — the other integrations are optional and each unlocks its own tool group. Copy the API key from Semust settings and put it in your client config as `SEMUST_API_KEY`, along with `SEMUST_BASE_URL`. Python 3.10+. There is no package install: download or clone the repository, `pip install -r requirements.txt`, and point your client at `semust_mcp.py` with python. The README ships an `alwaysAllow` list covering every tool, since all of them read.