On this page
A translation management system (TMS) is software that holds your translations in one place and runs the work around them: who translates what, review, translation memory, glossary, machine and AI translation, quality checks, and the way finished translations get back into your product or your documents. It replaces files passed around by email and spreadsheets with one workflow that keeps up as the content changes.
This guide covers what a TMS does, how it differs from a CAT tool and a localization platform, the three kinds of TMS you will meet, when you need one, how TMS pricing works and what to compare.
- What it is: one place for translations plus the workflow around them: assignment, review, memory, terminology, AI and MT, checks, integrations, delivery.
- vs CAT tool: a CAT tool is one translator's editor; a TMS organizes the work of everyone and connects to where the content lives. Most TMSs include a CAT view.
- vs localization platform: mostly naming; "localization platform" stresses what goes beyond translation (sync from code, review, delivery) and is what software-focused tools call themselves.
- Three kinds: project and document TMS, continuous TMS for software, enterprise suites.
- When you need one: more than two or three languages, several translators or reviewers, text that changes every release, or non-developers who change text.
- Pricing: per user, per hosted word or string, per key, per translated volume or per usage; the meter decides the bill more than the headline price.
What a translation management system does
| Function | What the TMS does | Without a TMS |
|---|---|---|
| Workflow | Assigns work per language, tracks progress, routes every change to review | Emails, spreadsheets, status meetings |
| Translation memory | Stores every approved source and target pair and suggests it for new text | The same sentence gets translated (and paid for) again |
| Glossary and style guide | Keeps product terms, forbidden terms and tone consistent in every language | Each translator decides alone |
| Machine and AI translation | Drafts translations, ideally with the glossary, style guide and memory as context | Copy and paste from a translation tab |
| Review and history | Records who changed or approved each translation, and when | Nobody can tell where a string came from |
| Quality checks | Finds broken placeholders and tags, length problems, untranslated or inconsistent text | Users find it after the release |
| Integrations | Syncs with code repositories, CI, a CMS or design tools over API and CLI | Manual export and import |
| Delivery | Puts finished translations into the product: as files in the build, or from a CDN at runtime | A release for every text change |
A good TMS treats a change like a code change: the new value waits for review, the old one stays live until someone approves, and the history shows both. That is the difference between a shared folder of translation files and a system you can rely on when ten people touch the same strings.
TMS vs CAT tool vs localization platform
| CAT tool | Translation management system | Localization platform | |
|---|---|---|---|
| Who works in it | One translator | Translators, reviewers, project managers | Developers, translators, reviewers, product people |
| Unit of work | A document or file | Projects and languages | Keys (strings) in a product |
| Main job | Translate a segment well, with memory matches and QA | Organize the work, memory and terminology of many people | Keep every language of a product in step with its code |
| How translations come back | A translated file | Files or connectors, per project | Synced to the repository, built in, or served from a CDN |
The lines are blurry on purpose: most translation management systems include a CAT view in the browser, and "localization platform" is usually what software-focused TMS vendors call themselves. If what you translate is a software product, see what a software localization platform adds on top of the basics above.
Three kinds of translation management systems
- Project and document TMS. Built around jobs: a file arrives, is split into segments, goes to translators and reviewers, and returns translated. It fits documents, marketing content and translation agencies, and often includes vendor and cost management.
- Continuous TMS for software. Built around strings with keys that change every sprint. New keys reach the TMS from the code (an SDK, a CLI step in CI, an API or a coding agent), translation runs in parallel with development, and translations go back into the repository or out through a CDN without a release. This is what continuous localization means.
- Enterprise suite. Connectors to many content systems, approval chains, vendor management, and private-cloud or on-premise options for large organizations; the enterprise TMS post covers what they add.
Many products sit between two kinds. Choose by what your content is and how often it changes, not by the length of the feature list.
When you need a TMS
You need one when files, spreadsheets and email stop keeping up. The usual triggers:
- more than two or three target languages;
- more than one translator or reviewer working in parallel;
- text that changes every release instead of once a year;
- terms that must stay consistent: brand names, legal or regulated wording;
- people outside engineering who need to change text without a pull request;
- a need to show who translated and approved what, for customers or for regulation.
You probably do not need one yet with one or two languages handled by one person who reads both, or content that changes a few times a year. Some teams also start with a coding-agent routine that drafts translations in pull requests; when that routine is enough and when it is not has the details.
How TMS pricing works
| Meter | What makes the bill grow | Usually cheapest when |
|---|---|---|
| Per user (seat) | Team size, including occasional reviewers | A small, fixed team |
| Per hosted word or string | Content times number of languages | Few languages |
| Per key | Number of keys, not languages | Many languages, stable key set |
| Per translated word, MT characters or AI tokens | How much new text gets translated | Little new text per month |
| Usage (requests, CDN downloads) | Traffic of your app | Low traffic, or translations bundled in the build |
Most vendors combine two or three meters, so compare a whole year of your own project, not the headline price. How to choose a TMS in 2026 prices one reference project at nine vendors and shows where each meter wins. Locize plans include fixed amounts of words, languages and CDN downloads (AI tokens on paid plans), from a free plan to $199 a month; from the Growth plan, extra capacity comes in blocks. The pricing page has the numbers.
How to compare translation management systems
- Your files and your stack. Does it read your formats (JSON, XLIFF, PO, Android XML, iOS strings, RESX) and sync with your repository or CI?
- Delivery. Translations in the build, or from a CDN at runtime so a text change needs no release? Can you choose per platform?
- Translation sources. AI and MT with your glossary and memory as context, your own translators, ordered professional translation, and a review step for each.
- Workflow at scale. Branches and versions for releases, roles per language, several products or customer-specific wording.
- The pricing meter. Which meter grows with your plans: more languages, more people, more content or more traffic.
- Records and exit. History per string, a record of who reviewed what, and an export of everything, memory and glossary included.
For vendors side by side on these points, see the best localization platforms for developer teams.
Where Locize fits
Locize is a continuous translation management system for software teams, built by the team behind i18next:
- New keys arrive from the code: through i18next
saveMissing, the CLI in CI, the API, or a coding agent over the MCP server. - AI translation with context and a score: automatic translation uses your glossary, style guide and translation memory as context, and each AI translation gets a confidence score; with review routing on, the doubtful ones go to review by themselves.
- Delivery without a release: web apps, React Native and other runtimes that load JSON get text changes from the CDN; compiled iOS or Android string files sync into the build with the CLI, since Locize has no over-the-air SDK.
- Human translation when you want it: through connected translation order services.
Every account starts with a 14-day full-feature trial without a credit card: connect your app and try it on your own strings. How this works for a whole software product is on the software localization platform page.
Frequently asked questions
What is a translation management system (TMS)?
A translation management system (TMS) is software that holds a product's or a company's translations in one place and runs the work around them: who translates what, review, translation memory, glossary and style guide, machine and AI translation, quality checks, integrations with code repositories or a CMS, and the delivery of finished translations. It replaces files passed around by email and spreadsheets with one workflow that keeps up as the content changes.
What does a TMS do?
Eight things: (1) workflow: assigning work per language, tracking progress and routing changes to review; (2) translation memory: reusing approved translations for new text; (3) glossary and style guide: keeping terms and tone consistent; (4) machine and AI translation: drafting translations, ideally with the glossary, style guide and memory as context; (5) review and history: recording who changed or approved each translation; (6) quality checks: placeholders, tags, length, untranslated text; (7) integrations: API, CLI, repository, CI and CMS sync; (8) delivery: putting finished translations into the product, as files in the build or from a CDN at runtime.
What is the difference between a TMS and a CAT tool?
A CAT (computer-aided translation) tool is the editor one translator works in: source and target side by side, translation memory matches and QA checks for the segment at hand. A TMS is the system around it: it assigns the work, keeps memory and terminology for every translator, connects to the places the content comes from and goes to, and records review. Most modern TMSs include a CAT view in the browser, so translators need no separate desktop tool.
Is a localization platform the same as a TMS?
Mostly naming. "Translation management system" is the older, established term; "localization platform" stresses that the tool covers more than translation: strings with keys synced from the code, translation in parallel with development, review, and delivery into the build or from a CDN. Software-focused tools usually call themselves localization platforms; tools built around documents and agency projects usually keep the TMS name.
When do I need a TMS?
When files, spreadsheets and email no longer keep up. Common triggers: more than two or three target languages, more than one translator or reviewer working in parallel, text that changes every release, terms that must stay consistent (brand, legal, regulated wording), or people outside engineering who need to change text without a pull request. With one or two languages, one person who reads both and content that rarely changes, files in the repository are often enough.
How much does a translation management system cost?
It depends on the meter more than the headline price. TMS vendors charge per user, per hosted word or string, per key, per translated word or AI volume, or per usage such as CDN downloads, and most combine two or three of these. Seat pricing grows with the team, word pricing with content times languages, key pricing with content only. Locize has a free plan and paid plans from $7 to $199 a month that include fixed amounts of words, languages and CDN downloads (AI tokens on paid plans); from the Growth plan, extra capacity comes in blocks.
What are examples of translation management systems?
Widely used translation management systems include Crowdin, Lokalise, Phrase, Smartling, Transifex, Tolgee and Locize. They differ less in their feature lists than in what they are built around: documents and agency projects, strings in a software product, or connectors for a large enterprise. Compare them on your file formats, your delivery model, the pricing meter and how review is recorded.
Can a coding agent replace a TMS?
For some teams, for a while. A coding-agent routine that drafts missing strings and opens a pull request works while the team reads every target language and nobody outside the repository changes text. Past that point the routine has to rebuild translation memory, an enforced glossary, review per language with a record and delivery without a redeploy. Most teams keep the agent as the drafter and let it write into a TMS that already has those parts.
Try a translation management system on your own strings
Connect an i18next app or upload your translation files, let AI translate with your glossary as context, and review the result. 14 days, every feature, no credit card.
Start the free trial
