The hard part of searching pre-modern Japanese is that the text does not agree with itself. Here normalisation is part of the query path rather than something you do first, and you can see what a term will be turned into before committing to it.
A search layer over the Genji API for classical Japanese literature. Historical texts spell the same word several ways, so the server folds that variation in — historical kana, dakuon, kanji-kana variation, phonetic change, repeat marks — and lets you see what a term becomes before you search with it.
- Full-text search across classical Japanese texts, scoped to a volume when you want it
- The normalisation rules available, listed so you can choose among them
- A preview of what a given string becomes under the current rules
- A health check on the Genji API
Npx, or a global install of @nakamura196/genji-mcp-server. Every answer comes from the Genji API at request time, so it needs network access; the documented configuration is the launch command alone.
One command — npx @nakamura196/genji-mcp-server