// the take
A Python MCP server wrapping RollingGo's (formerly DIDA) hotel search API, exposing three tools — searchHotels, getHotelDetail, getHotelSearchTags — so an LLM agent can query a 2M+ property inventory for availability and pricing. It's aimed at developers wiring hotel search into Claude/Cursor/Codex-style agents rather than anyone wanting to actually book a room through it.
The tool surface is small and sensible — three focused calls covering search, detail/pricing, and tag discovery, which keeps the schema an LLM has to reason about manageable. Streamable-http transport with a documented curl example makes it easy to verify the server works before wiring it into an agent. Local deployment via uv with a single requirements.txt is genuinely zero-friction for trying it out.
This is a thin client around a single proprietary backend you don't control: the 'local deployment' option is explicit that it just forwards every call to RollingGo's hosted API, so there's no actual local functionality or fallback if their service has an outage or changes pricing. Despite the description's 'booking rates' framing, the API-key tier only supports search and pricing queries — actual booking requires a separate OAuth enterprise flow gated behind a sales contact, so what's open-source here is strictly read-only. The free/unlimited pricing is tied to a partner-commission business model (markups, payouts), which means the incentive structure behind the data you're querying isn't neutral. There's no test suite in the repo, and with 85 stars and a handful of source files, this has had very little independent use or scrutiny — you're trusting a young project with your agent's travel-booking logic and an API key tied to a revenue account.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →