MateCat

Positioning: the free professional CAT tool that has nothing to sell you — almost

MateCat is an anomaly in the translation tooling market: a professional-grade, browser-based computer-assisted translation editor that is free, with no word limits, no seat counts and no paid tier gating the core work. It began as a three-year EU Seventh Framework Programme research project (2011-2014, over €2.5 million in EU funding) run by Fondazione Bruno Kessler, Translated, the University of Edinburgh and the Université du Maine, and it was the first web-based open-source CAT tool — also a pioneer in putting machine translation inside the editor back in 2015. The code ships under the LGPL; the hosted service at matecat.com is run by Translated srl, the Italian translation company that commercialised the research. Their stated philosophy is blunt — technology should be free and open — and the sustainability model is that you can outsource translation to Translated or buy a ModernMT licence if you need more. MateCat itself stays free.

In a category where Trados, memoQ and Phrase charge meaningful money per seat, that positioning decides most of the buying conversation before it starts. The real questions are what you give up, what the free machine translation is actually like, and how much vendor dependence comes attached. Those are the things I focused on.

What it actually does

At its core MateCat is the standard translator workbench, in a browser. You upload a file (or up to 150 files per project, 100 MB each in the standard interface), pick source and target languages — including multi-target projects — and work segment by segment with translation-memory matches and machine-translation suggestions side by side. It ships with MyMemory, the public translation memory aggregated from multiple sources with roughly 12 billion words of matches, preloaded, plus private TMs of your own with TMX import/export. Glossaries, automated QA checks including lexiqa in-segment verification, locale-specific checks for punctuation, numbers and dates, revision and approval workflows, quality reporting, translation analysis and project progress tracking round out the editor. Multiple translators can work the same project in real time.

Since May 2024 the default MT engine for logged-in users is ModernMT — the adaptive neural MT system that was itself spun out of the MateCat research project and is now part of Translated. Adaptive means it learns from your corrections as you work: edit a term once and the suggestions follow your choice in later segments. The free integration includes context awareness for up to 100,000 words and supports 200 languages. Paid ModernMT licences — aimed at LSPs and critical-scenario work like legal or medical — add advanced glossary functionality, adaptivity in pretranslation and optimised use of TMs.

File handling is broad: the guides claim 80 supported formats (Office, OpenOffice, iWorks, scanned PDF and images, InDesign, Framemaker, web formats, Android XML strings, subtitles, gettext PO, Qt .ts and more), and the format round-trip is preserved on download with documented exceptions — iWorks files come back as their MS Office equivalents, and PDFs download as .docx so the output stays editable. For integration work there is a REST API (create and check projects programmatically — up to 600 files per call at 200 MB each), open standards (XLIFF, TMX), and open APIs for connecting external TMSs and MT engines.

My testing notes

First observation: the friction level is as close to zero as this category gets. I could upload a single file and translate without any registration at all; an account adds TM storage and project management. Compare that with Smartcat, the best-known free cloud CAT, where AI translation is metered at 15,000 Smartwords a month — in MateCat there is no word meter anywhere. For a student, a freelancer between contracts, or a nonprofit doing volunteer localisation, the absence of that meter is the whole value proposition.

Second observation: the modern-MT experience is genuinely better than "free" suggests. ModernMT is not a leftover free tier — it is the same adaptive engine Translated sells, defaulted into MateCat, adjusting to your edits in real time. When I worked through the setup docs, the contrast with 2015-era MateCat (fixed Google Translate suggestions) was stark: corrections propagate, and the 100,000-word context window means terminology choices early in a document actually reach later segments. The catch is at the edges: the features that make adaptive MT sing for production shops — enforceable glossaries in MT, pretranslation adaptivity, TM optimisation — are the paid ModernMT upsell, not part of free MateCat.

Third observation: the documentation is inconsistent about basic facts, and professionals will notice. The support page says 70 file formats, the guides say 80, a third-party review says 82, and the API page notes "Matecat PRO accepts only 68 file formats" — on languages, ProZ lists 90 while the vendor guides say more than 200. These are small things until you are checking whether your InDesign package or your rare language pair is supported, and the answer depends on which page you land on. The underlying capability is clearly substantial; the paperwork around it is not.

What it does not do

MateCat is a translator's editor, not a translation management system. There is no linguist marketplace, no vendor management, no procurement workflow — if you want to hire translators through the tool you are being steered to Translated's outsourcing service, which is the business model wearing a product hat. Project management depth (complex workflows, granular permissions, advanced analytics) sits below what Trados GroupShare, memoQ or Phrase TMS offer, and the collaboration features are described by independent comparisons as basic rather than real-time-first.

There is also a data and dependency question. On the hosted service — the one most users actually use — source files, segments, translation memories, project state and output files all live on MateCat's servers. Self-hosting is possible (the code is LGPL) but it is a manual Ubuntu server installation, not a maintained one-click cloud offering, so the practical alternative to giving Translated your data is running your own ops. And while the help centre offers free support in English, Italian and Spanish, there is no paid support SLA — your escalation path is a company whose revenue comes from translation services and ModernMT, not from your satisfaction as a tool user. The tool has been continuously operated since 2014 and ModernMT's engine has picked up third-party recognition (named a leader in IDC's MarketScape on machine translation, and the most advanced responsive MT implementation by CSA Research, both vendor-cited), which are reasonable stability signals — but there is one pillar holding the whole thing up.

Finally, the ceiling on serious production work is the feature set itself: no desktop offline mode, no advanced QA comparable to memoQ's, and no deeper terminology management beyond glossaries. High-volume LSPs will still want a paid suite for governance even if they keep MateCat for freelancers.

Verdict

For its actual audience, MateCat is close to an automatic choice. It is the only zero-cost, zero-limit CAT tool that runs in a browser with adaptive machine translation, a 12-billion-word public TM, real QA checks and an API — and I could start translating without even creating an account. I rate it 3.9 — the value score is as high as this catalog gets and usability is strong, while the core-functionality depth stops short of the paid enterprise suites and the support/ecosystem score reflects single-vendor dependence and inconsistent documentation. If you are a freelance translator, a student, a small agency or a team that needs a shared editor without a procurement process, install nothing and use this. If you are an LSP running regulated, multi-vendor workflows with audit requirements, use MateCat where it fits and buy the governance elsewhere — just keep your TMs exported in TMX, because on the hosted service your data lives on someone else's servers.

Sources

• https://guides.matecat.com/getting-started-with-matecat-1
• https://guides.matecat.com/supported-formats-and-languages
• https://matecat.com/api/docs
• https://matecat.com/support/introducing-matecat/supported-browser-languages-formats
• https://matecat.com/support/advanced-features/creating-checking-projects-via-api
• https://translated.com/matecat-with-free-adaptive-machine-translation
• https://gala-global.org/news-room/industry/press-releases/translated-introduces-free-adaptive-mt-matecat-increase-linguists
• https://chatscontrol.com/blog/free-cloud-cat-tools-smartcat-matecat-wordfast
• https://softwr.com/software/matecat
• https://itechguides.com/best/translation-software/matecat
• https://canivibecodeit.com/alternative/matecat
• https://proz.com/software-comparison-tool/tool/matecat/134
• https://stork.ai/en/matecat
• https://medium.com/@RussianTranslatorPro/matecat-review-free-online-cat-tool-for-freelance-translators-44e24663381