Frequently asked questions
Short answers to what teams ask before they connect a repo. Each one links to the full answer in our docs.
Frameworks & compatibility
Which i18n libraries do you support?
next-intl, react-i18next / i18next and Lingui. When we set up i18n for you, we pick by stack: next-intl for the Next.js App Router, i18next for the Pages Router, and Lingui for React with Vite. If one of them is already wired up, we keep it and only translate.
Which file formats do you support?
JSON (flat or nested), .po, XLIFF 1.2 and 2.x, ARB, YAML, Apple .xcstrings, Android strings.xml and Chrome extension messages.json. Point the locale path pattern at your files and they stay translated on every push.
Does it work with monorepos?
Yes. Create one project per app and point its locale path pattern at that app's files, for example apps/web/messages/{locale}.json.
Does it work with mobile apps (iOS, Android, Flutter)?
Partly. We keep iOS (.xcstrings), Android (strings.xml) and Flutter (.arb) translation files translated on every push. Setting up i18n inside native or Flutter code isn't supported yet; that part works on React web apps.
Do you support browser extensions?
For translation, yes: Chrome messages.json and WXT locale files stay translated on every push. Setting up i18n inside extension code isn't supported yet.
Translation quality
How good are the translations?
Better than phrase-based machine translation, because the model sees the whole string, its placeholders, your glossary and your style guide at once. Cheaper and faster than a human translator, but not a replacement for native review where the stakes are high. A second AI pass flags anything that needs a human.
Can I use my own glossary and style guide?
Yes. Add protected terms under Project → Glossary (bulk import works) and set tone and formality per language under Project → Style Guides. Every job follows both.
How does QA work, and what does it check?
After translating, a second AI pass checks every language for meaning, placeholders and variables, glossary terms and style-guide fit. Results show on the project dashboard as good, passable or needs attention, and you can get a webhook or email when there are issues.
Can I edit translations after they're generated?
Yes. Change a translation in the pull request or later in your repo. Your edit becomes the approved version, goes into translation memory and isn't translated again.
Which AI models do you use? Can I choose?
Translation and review run on Anthropic's Claude models. The one-time code setup uses a separate coding model that moves up to a stronger one when a file needs it. We choose models for quality and cost; picking a model yourself isn't available today.
Workflow
How does the GitHub and GitLab integration work?
Connect the repo in the app. We scan it, set up i18n if it needs it, and push to a branch for you to review. After that, every push with new or changed text opens a pull request (a merge request on GitLab) with the translations. Azure DevOps is on the roadmap.
What triggers a translation job?
Three things, all settings you control: pushes to watched branches, pull requests if you turn on Translate pull requests, and manual jobs from the dashboard. A push without text changes doesn't start a job.
Can I translate specific branches, not just main?
Yes. Watch as many branches as you like, or run a manual job on any branch and choose between a pull request and a direct push.
How do I review translations before they merge?
Like any other code change. Read the pull request, edit a translation in the diff if you disagree, and merge when you're happy. Nothing reaches your default branch without a pull request, unless you pick direct push on a manual job.
What does the pull request look like?
One pull request per job. It only touches your locale files, only adds new or changed keys, and leaves translations you've already reviewed alone.
Team use
Can multiple people use one account?
Yes. Invite teammates by email under Settings → Users. No plan charges per seat.
How do permissions work?
There are two roles today. Admins invite and remove people, change roles and manage billing; Members work on projects. Finer-grained roles and SSO are on the Team plan roadmap.
Pricing
What counts as a "character"? Source only, or translated too?
We don't bill characters, keys or seats. You pay for text translated into each target language, quality review included: about €0.10 per 1,000 translated words, or roughly €0.015 per 1,000 characters. Your source text is never billed on its own.
How much does a typical project cost per month?
A small app with 500 strings in 3 languages costs about €0.90 to translate the first time, and cents a month after. A mid-size product with 3,000 strings in 8 languages is about €16 up front and around €1.60 a month. Repeated strings come from translation memory and cost nothing.
Indie, Starter or Team: which do I need?
Indie (€5 a month plus usage) is cheapest while you translate less than about €15 a month. Past that, Starter (€20 a month with €20 of credit included) costs less. Team (€249 a month with €200 included) is for organisations that want personal onboarding and team features as they ship.
Is there a free trial?
Yes. Every new account gets €5 of translation credit for 14 days, no card needed. That covers roughly 50,000 translated words.
Security & data
Where is my code and data stored?
Your translations live in your own repository. On our side, projects, glossary, style guides, translation memory and job history are kept on Hetzner servers in Falkenstein, Germany. The working data of each translation job is deleted 30 days after the job finishes.
Do you store my source code?
Translation jobs only read your locale files. The one-time code setup clones your repo into an isolated container that is destroyed automatically when the run ends. We keep the resulting diff so you can review what changed.
Is it GDPR compliant?
globalize.now SIA is an EU company registered in Riga, Latvia, and our servers are in Germany. We don't offer a signed data processing agreement yet. Ask in the support chat or email [email protected], and we delete your organisation's data within 3 business days. SOC 2 is on our roadmap for Q1 2027.
Do translations go through third-party APIs?
Yes, to AI models, and none of them may train on your text or code. Translation and quality review go directly to Anthropic's Claude models. The one-time code setup runs DeepSeek models through OpenRouter with zero data retention enforced. The full provider list is in our docs.
Switching tools
I'm on Lokalise, Crowdin or Phrase. How do I switch?
Export your translations from the old tool into your repo, connect the repo, and set Existing translations to Import as already approved. Import your glossary too, and cancel the old tool once the first pull request lands.
Can I import existing translations?
Yes, in one of three ways: import them as already approved (added to translation memory, never translated or billed again), import them for review, or ignore them and translate from scratch.
Will it break my current i18n setup?
No. On an app that already uses next-intl, react-i18next, i18next or Lingui, we don't extract strings, restructure keys, add a language switcher or touch routing and config. We connect to the files you have and keep them translated.
Didn't find your question? Search the docs or email [email protected].