Including the ones where the answer is no. If something here rules us out for you, we would rather you found it on this page than three weeks in.
A translation management system is where the strings in your software live between the developer who writes them and the person who translates them. It holds one copy of every key, a column per language, who changed what, and an API so your build can read the finished files. locoloop is one, built for product teams rather than for a localization department.
Files in a repository work until a second language and a non-developer are involved. Then a translator needs a way to edit without a pull request, you need to know which strings changed since they last looked, and someone has to notice that a key was deleted. That is the whole job locoloop does.
Their override stands and is flagged as stale so a reviewer can decide. Cells that were inheriting update immediately.
No. Inherited cells are not counted twice — only the keys you actually override add to your total.
Yes. Invite them as editors on one project or one platform, and turn AI drafting off for any language you want handled by people. Reviewers and viewers are free and unlimited, so a colleague who speaks the language costs nothing to add.
CLDR forms per language, so Polish gets one, few, many and other, and Arabic gets six. An export refuses to ship a partial set, and the status endpoint names the forms a language is still missing.
Add German twice with different registers. They are independent languages, and the informal pass starts from your approved formal wording so the two differ only in the pronoun. Thirty-five languages carry their own rule.
JSON, Apple .strings and .xcstrings, Android strings.xml, XLIFF, gettext .po, YAML, Flutter .arb, CSV and Java .properties. Import what you have, export what your build already reads.
It is a parser, not a rewrite. Send us a sample file and we will add it.
Not in the classic sense. There is no fuzzy-match TM that proposes a previous segment at 85% similarity. What there is instead: a glossary that every AI draft honours, per-project context, and variants that inherit rather than duplicate. If a fuzzy-match TM is a requirement, say so before you start.
Neither today. It is a hosted service running in the European Union. Your data is exportable in full at any time, in every format we read, which is the part self-hosting is usually asked for.
In the European Union. The database runs in Frankfurt and backups stay in the EU. On the Business plan you connect your own AI provider key, so your text is processed under your own agreement with that provider.
Export everything in any format at any time, including full history as JSON. No egress fee and no export tier.
The fastest answer is your own strings in the editor. It takes one import.