Over the past year, AI coding agents have become incredibly good at localization. Claude, Codex, Cursor, and similar tools can generate strings, update components, and create translations in multiple languages from a single prompt. For many React applications, generating translations is no longer the difficult part.
The interesting problem now is everything that happens after the translations are generated.
Product managers still want copy changes. Translators still need review workflows. Teams still need context for each string. Developers still need synchronized translation files. The AI can generate the content, but somebody still needs to coordinate the workflow around that content.
While building i18nexus, we started wondering what would happen if AI coding agents created strings, translations, and context through the same system as everyone else instead of writing directly to translation JSON files. The result is a workflow where AI agents can generate strings, translations, and context through MCP, while PMs and translators review everything in a shared dashboard. Translation files stay synchronized automatically, and developers work with AI coding agents the same way they already do in their application.
The easiest way to get started is to run the init
command at the root of your project. It doesn't matter if you're using next-intl, i18next, or react-intl:
npx @i18nexus/cli init
The CLI creates an i18nexus project, imports your existing translations, configures synchronization, and sets up the MCP integration for Claude Code or Codex. No code changes needed. Continue working as you normally do.
Once configured, AI agents automatically add new copy and translations to i18nexus as you code, which are then synced down to your local JSON files. The developer experience doesn't change at all.
Those changes immediately become available for review inside the dashboard, where PMs and translators can edit and verify translations before they are synchronized back to local translation files.
Whether a PM edits a string in the dashboard or another developer adds a string from their MCP, the local translation files update automatically for all developers. The application updates. The build stays synchronized. Nobody, including AI agents, manually edits JSON files. We often find ourselves gitignore-ing them. They're now simply generated output.
The most interesting thing about AI localization isn't that AI can generate translations anymore. That's already becoming expected. The interesting challenge is figuring out how developers, translators, product managers, and AI agents work together around those translations.
For us, the answer has been to treat AI agents as first-class participants in the localization workflow rather than tools that edit translation files directly. The coding experience doesn't change, but the collaboration experience changes dramatically.
Check out the quick setup guide [here](https://i18nexus.com/getting-started).
Happy coding!