Coordinate translators and
track progress at a glance

Yes. Strings carry translation states — AI-translated, human-edited, and reviewed — and on top of that there is a real review and approval workflow: request review on a string, approve it singly or in bulk, and every approval is attributed to the reviewer who made it. Threaded comments with @mentions sit on the row itself, and a mention fans out to the mentioned user’s notification inbox, with daily or weekly digests and an optional Slack channel. Version history and revert are there too, so an approval is never a one-way door.

Invite your team into the organization and everyone works in the same shared editor. Four organization roles — owner, admin, developer, translator — gate what each person can do, and on top of that you can grant access per project and even per language, so a French reviewer can be given exactly the French rows and nothing else. That granular access control is included on every plan, Free included.

Exact-match memory matches on the source hash, target locale, and model, so once a given string has been translated into French it is reused instantly the next time that exact string appears — at no quota cost. Be aware the exact-match pool is a single global, anonymous one rather than a private per-organization store: entries carry no project or organization, so a match may come from elsewhere on the platform, and your translations can serve someone else’s exact match. Nothing records who contributed an entry. If you would rather not participate, switch translation memory off for the project. On Team plans and up there is also a semantic (fuzzy) memory — but that path is the opposite: it is scoped to your own organization and only surfaces near-matches from your team’s human-reviewed translations, as suggestions.

Quality Mode switches AI translation to Claude Sonnet 5 with a corrective retry pass, for strings where quality matters more than speed. It’s available on Team and Business plans and can be scoped to specific locales.

A glossary (Indie and up) lets you mark mandated per-locale terms, do-not-translate terms, and forbidden terms. Terms ride into the AI prompt and are checked deterministically after translation — warn mode flags a miss, enforce mode makes it a hard failure with a corrective retry. You can also build an org-wide house glossary and link it to every project.

Translation QA (Team and up) runs a free structural pre-gate — placeholders, plural completeness, length, terminology — then an AI review that scores each translation 0–100 with categorized findings. Anything that scores low or has a major finding is flagged needs_review, so you can triage instead of re-reading everything. Confidence-gated auto-approve takes it further: the model self-reports a confidence at translate time, and only structurally-clean, high-confidence rows auto-approve, so your team reviews the risky strings rather than all of them.

Yes — export the per-language MQM scorecard as CSV. Each locale rolls up into a score trend, its top finding categories, and the share of rows still needing review, so quality becomes a concrete artifact you can hand over. It is built from indexed reads at no extra AI spend, on Team plans and up.

Yes, with a per-project AI Profile (Team and up). Capture your brand voice, tone, and a few worked examples, and every AI translation for that project is grounded in them — alongside your glossary terms and the style guide transglot can infer from your reviewed translations.