Localization leads can compare how the same product vocabulary trends on Google Search, Google Shopping, and YouTube without rebuilding a custom data warehouse.
Teams ship perfect translations, then discover the category moved under different slang. Traditional rank trackers show positions only after assets exist. Trends MCP answers whether demand curves for the chosen phrases are rising or falling before copy ships, using the same google search and google shopping sources global SEO already cites in English language playbooks.
Rows are market intent phrases, not single dictionary outputs. Columns are sources: google search for broad demand, google shopping for purchase language, youtube for education first queries, and amazon when the catalog strategy spans marketplaces. Get Growth calls with identical percent_growth arrays make quarterly reviews legible to finance, because every cell used the same 12M or YTD definition.
In some regions Shopping reflects retailer ads and merchant density more than organic web blogs. In others, Search still carries informational mixed intent. Seeing both curves for the same stem phrase helps PMs decide whether a landing page should be editorial or transactional. Trends MCP keeps those pulls in one JSON contract so engineers do not maintain two parsers.
Global portfolio views already have a home at https://www.trendsmcp.ai/global-seo-trend-research. Core keyword framing sits at https://www.trendsmcp.ai/seo-keyword-research. When merchandising also wants proof, add https://www.trendsmcp.ai/google-shopping-trends and https://www.trendsmcp.ai/amazon-search-trends. YouTube led markets should cross link https://www.trendsmcp.ai/youtube-trends.
Homonyms break charts. Document disambiguation rules in the keyword sheet, for example always pairing the brand with the product noun in markets where the brand reads as a common word. When the API returns not_found, treat that as a vocabulary bug, not a tool failure.
Use the documented prefixes from the product prompt appendix so models call the server instead of inventing numbers: "Using TrendsMCP, compare 6M Google Search growth for [phrase US] vs [phrase DE]." Follow with "Via TrendsMCP, show Google Shopping trends for the same pair." Consistent phrasing matters more than clever prompt engineering here.
POST bodies follow the public reference on https://www.trendsmcp.ai/docs. Weekly jobs can hit the same endpoints analysts use inside Cursor, which keeps QA and production aligned. Always send the Authorization header server side only; never embed keys in front end bundles.
Each source plus keyword pair consumes a request on Get Trends or Get Growth per the billing overview in llms.txt. A ten phrase matrix across four sources is forty calls before any retries. Plan paid tiers when the program covers more than a handful of countries, and batch pulls to stay inside monthly caps on the free tier during pilots.
Trends MCP does not return SERP positions, hreflang tags, or crawl depth. It returns demand and attention style signals. Say that plainly in slide footers so localization PMs do not over interpret a single spike around a celebrity name collision.
FAQ