Using a coding agent

Localizer’s integration instructions are written for coding agents as well as people: AGENTS.md, the AGENTS.md at the root of the repository. It walks through detecting the language and framework, adding the dependency, creating the locales package, adding the one line, hooking the CLI’s own output helpers, pinning tests to English, verifying with pseudo-localization, and connecting the repository. An index of these pages for agents is at llms.txt.

Give your agent a prompt such as:

Read https://locale.dev/AGENTS.md and integrate Localizer into this CLI. Add the GitHub Action workflow it describes.

Then review the diff like any other change: a dependency, a locales package with empty catalogs, one line before the app runs, a few T/t calls at output chokepoints, an English pin in the tests, .localizer.yml and a workflow. Once merged (and the repository is connected), the first Localizer run opens the pull request that fills in the catalogs.

The agent doesn’t translate anything itself: translations always come from the service, and your corrections to them are never overwritten.