Reporting and writing are separate halves of the same set: coverage, quality and usage analysis say what is absent, and the sync tools then add those keys from a source language in sorted order, with a backup available and a dry run on the bulk paths. Because the scanner reads source code for keys that never reached a file, drift shows up from both directions. Translation itself stays out of scope — the sync tools copy values across or drop in a placeholder, leaving the wording to whoever does the language work.
A server over an i18next project's locale files that reports on their coverage and quality, and writes missing keys back into them.
- A health check across the translation files, with per-language and per-namespace detail, and JSON syntax and structure validation that can fix what it finds.
- Coverage and quality reports by language and namespace, plus a listing of every translation file.
- Key usage analysis against your source paths, and a scan that finds keys used in code with no entry in the files.
- A detailed list of missing keys by language and namespace, in a format you choose.
- Sync paths that fill the gaps from a source language — named keys, chosen namespaces, or everything missing at once — each able to take a backup first, with a dry run available on sync_missing_keys and sync_all_missing.
- A single-key write that adds a key with its values to named languages and namespaces, and an export of translation data in various formats.
An i18next project on disk, and four environment variables telling the server where it lives and what it covers: I18N_PROJECT_ROOT, I18N_LOCALES_PATH, I18N_DEFAULT_LANGUAGE and I18N_SUPPORTED_LANGUAGES.
One command — npx -y i18next-mcp-server@latest