Use cases / Marketing and web

How we translate this website with Mondoword

Every string on this site is a key in our own Mondoword project. Here is how that works, and what we are still building.

One table for all target languages, with progress per language

We build this website the way we would like our customers to build theirs: no text is hard-coded in components.

How it works

  1. One message file. Every string on the site lives in messages/en.json, with keys such as hero.subtitle and faq.items.0.q.
  2. One project. The same keys live in a Mondoword project, “Mondoword website”, with English as the source language.
  3. Designs linked. The website designs in Figma are linked to the same keys, so the AI translates headlines and buttons with the page they belong to.
  4. Sync with a script. npm run messages:push creates keys for new strings in en.json; npm run messages:pull <lang> writes messages/<lang>.json from the project. Strings without a translation stay in English.
  5. Translate, then publish. When we add a language, we translate in Mondoword, review the drafts, pull the file and publish it.

What we are learning

Translating our own site shows us what web teams need next. On our list:

  • a real command-line tool and CI step (today it is a small script and a personal token),
  • plurals and placeholders (ICU messages),
  • long-form pages such as help articles, which are documents rather than short keys,
  • right-to-left layouts: the plugin already shows Arabic and Hebrew right to left, but the canvas preview keeps each layer's alignment.

FAQ

Why is the site only in English?

We launch in English and add languages once each one has been reviewed. Search engines treat large amounts of unreviewed machine translation as low quality, and so do readers.

More use cases