zhtw-mcp
A linguistic linter for Traditional Chinese (zh-TW) that plugs into AI coding assistants over MCP and enforces Taiwan Ministry of Education standards on vocabulary, punctuation and character shapes, catching Mainland Chinese regional drift.
The tool enforces three official Taiwan standards: the Revised Handbook of Punctuation for punctuation marks, the Standard Form of National Characters for character shapes, and cross-strait vocabulary normalization grounded in OpenCC's TWPhrases and TWVariants datasets for word choice. Over 1,100 vocabulary rules and 15 casing rules are compiled directly into the binary.
For an ambiguous term the rules do not resolve, the server asks the AI assistant it runs inside for a decision, so no separate API key is needed. The project's motivation is that AI models trained mostly on Simplified Chinese web text tend to default to zh-CN terminology even when asked to write zh-TW, in ways that go beyond character conversion, since the same word can mean different things on each side of the strait, such as 並行 meaning "concurrent" in Taiwan but "parallel" in the mainland.