Quetzal

The core of Quetzal is translation that reads the room. Instead of sending isolated strings to a model, it gathers the code around each one, the components and props nearby, and comments that explain intent. That context tells the model whether it is handling a button, a heading, or an error, which is where machine translation usually falls apart. The limit is that the quality of the surrounding code matters. A clean, well-structured repository gives it more to work with than a tangled one.

Brand names, product nouns, and domain jargon get a glossary that Quetzal enforces instead of guessing. It also reads your existing translations to infer tone and register, then holds that voice steady across every locale and release. For teams that care about a consistent brand voice in ten languages, this is the part that replaces a style guide nobody reads. It works best when your existing copy is already consistent enough to learn from.

Quetzal does not just translate. It can audit work another tool already did, flagging errors and proposing fixes, and it scans for user-facing text that never made it into localization. That second job matters more than it sounds: hardcoded strings are how a perfectly localized app still shows raw English in a dialog box. Both audits run as part of the same review-and-merge loop, so nothing extra gets bolted onto your workflow.

Quetzal is not another translation library to adopt. It plugs into the stack you already run, with first-class support for JavaScript, TypeScript, React, Next.js, Swift, Python, Kotlin, Java, HTML, and XML, and it works with any i18n framework. That means no rewrite and no pipeline migration. The catch is that it still needs a repository to link. Projects without a codebase it can read, or teams that keep translations outside version control, are a poor fit.

Updates arrive where developers already work. Quetzal opens pull requests or makes commits, so translations go through the same review, history, and rollback as any other code change. There is no separate dashboard to check and no export-import dance. For teams that live in git, this is the whole pitch: localization becomes a background process instead of a recurring chore.

Static strings in a repo are one thing, but user-generated and runtime-generated text is another. Quetzal offers an API (hosted at api.qtzl.ai, authenticated with an API key) for translating non-static content. The company asks teams to talk through the context profile first, so the API fits content that actually benefits from it. This is the piece that pushes Quetzal beyond a code-only tool. It is still aimed at developers rather than casual users.