The engine, as tools.

Other agents can call the same tools Oraqul uses. Each call returns its data and a receipt for every Qloo call it made.

Run it

The server speaks MCP over stdio. It runs on demo data unless you set QLOO_MODE=live and a Qloo key.

pnpm --filter "@oraqul/mcp..." build
node packages/mcp/dist/stdio.js

What a call returns

Every tool returns its data and one receipt for each Qloo call it made, as JSON text. A score, a verdict or a confidence in the data was computed by code from what Qloo returned. The tools send only public names, place names and age bands to Qloo.

Tools

  • searchEntityFinds a title, artist, brand or place in Qloo by name.
  • resolveBundleDescribes a launch as a bundle of comparable titles, one exact match each.
  • audienceBySegmentAge and gender skews for an audience, as an index per segment.
  • rankPlacesRanks cities by audience strength from one Qloo call.
  • liftScores candidates against a baseline from one ranking call.
  • scoreRivalsHow much of an audience each rival release shares, in one call.
  • findTasteWhat an audience likes in one kind of entity, shaped by place and age. Brands are found by category and city, then ranked as one list.
  • partnerFitBrands, podcasts and series a film or title audience also likes, ranked by Qloo for one signal. A plan finds brands with findTaste by category and city instead.
  • topicMapThe topics an audience carries that rival shows’ audiences do not.
  • resolveTagsTurns a phrase into a Qloo tag, or null. Never a guess.
  • themeTagsThe theme and archetype tags that best describe a set of titles.
  • filmProfiles, bookProfiles, gameProfilesWhat Qloo lists for each title. Nothing is scored.

Start anywhere.
Leave with a plan you can revise.