PinPinterest — Pinterest Keyword Research

Each tool targets a distinct slice of the domain: single-keyword metrics (keyword_lookup), winning pins for a keyword (keyword_top_pins), winning pins for a niche (niche_top_pins), co-occurring keywords (related_keywords), rising topics (trending), plus two utilities and a queueing call. The only mild overlap is the keyword_top_pins / niche_top_pins / trending cluster, where all three return pin lists and an agent must read the descriptions to pick correctly. That overlap is small and well-explained, so misselection risk is low.

All names are snake_case and mostly descriptive, but the conventions are mixed: verb_noun (list_niches, request_keywords), resource-prefixed noun phrases (keyword_lookup, keyword_top_pins, niche_top_pins), and bare/possessive names (trending, my_quota). There is a loose resource-prefix grouping (keyword_*, niche_*) that helps readability, but it is not applied consistently across the set. Readable, not predictable.

Eight tools is well-scoped for a keyword research server: one lookup, three discovery/listing tools, one queueing tool, and two free metadata utilities. No tool looks redundant or like filler, and the surface is small enough for an agent to hold in context.

The lifecycle is covered end to end: check a keyword, get winning pins, expand to related keywords, browse niches and trends, queue missing keywords for crawl, and monitor quota. The main gap is the lack of a bulk lookup for several already-tracked keywords in one call, which forces the agent into repeated single-keyword calls (bulk only exists on the request/crawl side). That is a workaroundable inconvenience rather than a dead end.