https://open-gpt.app/ VS Currawong Web

Currawong Web's answer:

Overseas buyers and sourcing teams who need to check a Chinese supplier and cannot reliably reach the Chinese official portals themselves — importers placing a first order, procurement teams onboarding a new factory, and small distributors without a China office or Chinese-language staff.

It suits one-supplier, one-decision checks rather than continuous screening at scale. A team already running a subscription database and reading Chinese records in-house does not need us.

Currawong Web's answer:

It is staffed from inside China, and it publishes its own limits.

Every delivered line names the platform it came from and the date it was queried, because a Chinese registry record is a dated fact and stops being current the moment something is filed. Person names are withheld by default and unlocking them is a separate flat fee, so a routine supplier check does not quietly become a file on individuals.

The part most buyers never hear about: the registry API this industry runs on does fuzzy matching. Ask for a company it cannot find and it returns the closest one it has — silently, with a complete-looking record. We reject any result whose registered name is not character-for-character the one ordered, and say so, rather than delivering another company's records under your supplier's name.

Currawong Web's answer:

Because you can check our work before you pay us anything.

The registry snapshot — scope, status, capital, established date, address — is free. Two tools run entirely in your browser with no account and no upload: a factory-or-trader read of a registered business scope, and a Unified Social Credit Code checker that validates the check character against GB 32100-2015.

Beyond that, our measurements are published as open datasets under DOIs, with methods and limits stated, including one conclusion we withdrew after a paired control experiment showed our first reading was wrong. Anyone can download the data and judge whether we are careful. That is a different offer from a report you can only assess after buying it.

Currawong Web's answer:

It began with a measurement rather than a plan. We were testing whether China's official company registry could be reached, and the first reading looked clear-cut. A paired control — same machine, same moment, the only variable being whether the request went through a proxy — showed the failure was in our own network path, not at the target, including for an Australian government host that answered fine on a direct connection. So we withdrew the conclusion and published the correction.

That set how the desk works: state the vantage point, state the date, state what the observation cannot support. Buyers who cannot check China-side records themselves are exactly the people who cannot audit an overconfident answer, so the limits have to travel with the figures.

Currawong Web's answer:

A static site with serverless functions and a serverless SQL database at the edge, with no third-party runtime dependencies in the request path. The two public tools are plain browser JavaScript — nothing is uploaded and no account is created to use them.

Registry data comes from one licensed Chinese business-information platform. We do not query, proxy, embed or scrape the official government portals from a buyer's browser, and we do not work around access controls on them; when an official source has to be read, a person in China reads it and reports the source, the query date and the limits.