MarketSense AI — a Qloo-grounded cultural discovery agent
Problem
Early-stage founders often position products using intuition, broad demographic stereotypes, or generic AI recommendations. MarketSense helps a founder turn one concrete cultural interest into verifiable Qloo evidence and a small set of measurable market experiments.
What the app does
- Founder enters a business concept, city, audience hypothesis and specific taste cue (for example coffee, jazz or hiking).
- A server-side tool first searches the Qloo Taste AI API's
/v2/tagsendpoint for an actual tag matching that cue. - Only after resolution does the agent request
/v2/insightsfor localurn:entity:placerecommendations, conditioned on the real Qloo tag ID and the chosen city. - The interface renders the returned entity names, their source ranks, an API affinity value where present, and a transparent tool trace.
- A deterministic planning step proposes reversible experiments: customer interviews, landing-page positioning tests, and potential partnership research tied to the returned cultural entities.
Distinctive point: An uninterpreted location-based list isn't enough. The second Qloo call depends on the tag discovered in the first, and the recommendations shown to the user depend on Qloo's actual output.
Architecture
- React + TypeScript interface; Zod validation at the server boundary.
- Floot-hosted server endpoint
POST /_api/analyze. - Qloo hackathon API base:
https://hackathon.api.qloo.com. GET /v2/tags(cultural cue resolution) →GET /v2/insights(taste-conditioned local place discovery).QLOO_API_KEYis read only on the server; never included in the browser or GitHub.- Transparent 400/422/502/503 states; no fabricated or offline substitute Qloo entities.
- No automated customer outreach, purchasing, or sensitive personal-data collection.
How to try
Open https://marketsense-ai.floot.app . Try a specialty coffee shop concept in Almaty with the cultural cue coffee, then click Run live Qloo discovery. With the credential connected, the page resolves a real Qloo tag and returns place suggestions. If a key is missing or the API request fails, the app shows an explicit error rather than invented results.
Development and verification status (30 October 2026)
- Public site deployed and latest code published.
- TypeScript typecheck passed and negative-path endpoint tests returned expected HTTP 400/503.
- The Qloo API credential connection is pending; successful live Qloo results have not yet been verified.
- Public GitHub repository has been created at https://github.com/daryas2226-rgb/marketsense-ai, but source sync has not been completed due to GitHub integration write permissions; do not treat it as a complete source repository until files are uploaded.
- Final competition submission not yet confirmed.
Limitations and ethics
Qloo's taste affinities are aggregate signals and do not prove individual preference, product-market fit, causal impact, local availability, or sales. The target audience is a hypothesis for research, not a Qloo demographic claim. Unknown API coverage can produce zero results. This prototype does not include accounts, persistent user data, or guarantees suitable for production use.
Links
- Web application: https://marketsense-ai.floot.app
- Source repository (sync pending): https://github.com/daryas2226-rgb/marketsense-ai
- License: MIT in the Floot source, planned for the public repository
Built With
- ai-agent
- cultural-intelligence
- floot
- qloo
- react
- typescript
- zod