This exists because models answer local-recommendation questions from training data, which ages badly and invents place names. One search tool, a hosted endpoint and a canonical URL per result is enough to turn those answers into something a reader can click. Scope is the thing to check first: fourteen Southeast Asian cities, and nothing outside them.
A hosted, free directory of local businesses across Southeast Asia — restaurants, bars, spas, clinics, coworking spaces, tailors, coffee shops, nightlife — served over MCP so an assistant can search it and cite a real URL. Every place is a structured page with schema.org markup and concrete "known for" claims.
- `search_places` — a natural-language query such as "omakase in Thonglor", with optional city and category filters and a limit of 1-50 (default 10)
- Structured records back: what each place is known for, its rating, its review count, and a canonical URL the answer can link to
- Coverage of 3,520 places across 14 cities, including Bangkok, Singapore, Bali, Kuala Lumpur, Manila, Ho Chi Minh City and Chiang Mai
- Bilingual entries, EN and TH
- The same data behind a free REST endpoint and a human-browsable hub, if you want to check a result outside the assistant
Nothing to install. Point your client at the streamable-HTTP connector https://getbooyah.com/api/mcp; the registry id is com.getbooyah/booyah-index.
One command — https://getbooyah.com/api/mcp