public-apis-cn: a Chinese language directory of public APIs

A curated list, not a library

The whole repository is a README file and a licence file, and that is the honest description of what you get. The project is the public-apis list translated into Chinese, with the advertising stripped out, plus a collection of APIs that are commonly used inside China. The stated ambition is to become the Chinese language free API encyclopedia, and the invitation at the top is to star the repository and to contribute a useful API. The value of such a list is real but narrow. It answers the question of what exists, and for a developer who does not read English it answers it in a language where the descriptions of domestic services exist at all, since the services themselves document in Chinese. What it does not do is tell you how to use an API, what its rate limits are, or what it costs at scale. For that you still go to the provider, and the list is a map rather than a manual.

The four columns that make the list usable

The entry format is a table, and the columns are the same across sections. The first is the API name, linked. The second is a description, which is editorial and informal in tone rather than technical. The third is the authentication method, and this is the column that saves the most time, because it tells you at a glance whether you need an apiKey, whether the service uses OAuth, whether it authenticates with a User-Agent header, or whether it needs nothing at all. The fourth is whether the service supports HTTPS, which decides whether you can call it from a modern browser or from a client that refuses plaintext. Some sections add a fifth column for cross-origin support, marked as CORS, and its values are frequently unknown, which is itself informative: it means nobody has recorded whether the service allows a browser page to call it cross-origin, and you have to find out by trying.

How paid and free are marked, and what the AI section says about free

Each description is prefixed with a marker that distinguishes a paid API from a free one, and the project explains the convention in the introduction. The markers are not precise about pricing, they are a rough orientation for the reader deciding where to look first. The AI models section is where the project becomes genuinely opinionated, and it is worth reading as a position rather than as reference data. The note says that AI models need strong GPU compute, that hardware investment and electricity costs are high, that they are usually charged, that free allowances are limited, often something like ten to twenty units of API credit with an expiry date, and that they are generally only for testing. It then makes the point that genuinely free access comes from open-sourced trained models, and names DeepSeek, Qwen, KIMI, ChatGLM and MiniMax as the domestic examples. The same section lists local runners such as Ollama and LM Studio, which need no authentication and do not support HTTPS because nothing leaves the machine.

Sixty categories, from animals to zero trust

The scale of the catalogue is the other thing to understand. There are around sixty sections, and the table of contents lists them all, from AI models and weather at the top through domestic logistics, international logistics, video, vehicles, URL shorteners, transport and tracking, text analysis, test data, sports and fitness, social, shopping and security, then science and mathematics, programming, photography, phone, personality, patents, open source projects, open data, news, music, machine learning, jobs, health, government and geocoding, on through games and comics, food and drink, finance, events, environment and entertainment, email, documents and productivity, dictionaries, development, data validation and currency exchange, then cryptocurrency, continuous integration, cloud storage and file sharing, calendar, business, books, blockchain, authentication and authorization, art and design, anti-malware, anime and animals. Two observations follow. The categories are the vocabulary of a working developer rather than a taxonomy, which makes the list easy to browse. And the domestic Chinese services sit alongside the international ones in the same tables, which is the addition the translation itself could not have provided.

Weather, as a worked example of the entry style

The weather section shows the format at its most detailed, with five columns because CORS is recorded there. Thirty entries run from 7Timer, which the description notes is slow from China but needs no API key, through the domestic services such as Moji Weather with minute and kilometre level forecasts, AMap weather for real time and multi day forecasts, QWeather and the Xin Zhi weather API, and out to international ones like AccuWeather, Aemet, AviationWeather, the Hong Kong Observatory, Meteosource, the Met Office DataPoint service, Open-Meteo, openSenseMap, OpenWeatherMap, QWeather's developer endpoint, RainViewer, Storm Glass and the US national weather service. Reading down the authentication column you can see the whole range at once, apiKey for most, OAuth for Foreca, User-Agent for the Norwegian meteorological institute, and no authentication for 7Timer, AviationWeather, the Hong Kong Observatory, ODWeather, Open-Meteo, openSenseMap, RainViewer and the US service. That column is where the practical decisions are.

What the project is not, and how to treat it

Two honest caveats sit under the usefulness. The first is that the entries are links and short descriptions, so a description can be out of date, promotional, or simply wrong, and a project with a single README has no automated link checking to catch a dead URL. The second is that the CORS column is unknown for a great many services, which means the most common reason a browser application fails against a public API is not answered by the list at all. The repository also has no releases and no tooling beyond the README and the licence, so the freshness signal is the last push date and the diff of the file. The Apache-2.0 licence covers the list itself, which means you can fork it, translate it further or fold the entries into your own internal catalogue, and that is probably the most likely way a team would use it: as a starting list that someone in the team then verifies against the services they actually depend on.

Editorial conclusion

The directory is worth bookmarking if you build for a Chinese audience and need to find a weather, logistics, currency or AI API without reading English documentation, and the authentication and HTTPS columns save you a lot of trial and error. It is the wrong tool for a production dependency list, because the entries are links and short descriptions rather than specifications, and the cross-origin column is unknown for many of them, which means you still have to test from a browser. Before you rely on any entry, check the current terms of the service yourself, and remember that the repository holds a single README with no releases, so the last push date of 2026-10-15 is the only freshness signal you get.

Frequently asked questions

Official sources

1. Issues
2. License: Apache-2.0
3. llf007/public-apis-cn on GitHub
4. README