Tolgee

Tolgee is an open-source localization platform, or translation management system, built for software teams rather than translation agencies. It gives you a place to store every translatable string in your product, a web editor where translators work, and a set of SDKs that connect your running application back to that store. The core idea is that you should never have to hunt through source files for keys, hand-edit JSON or PO files, or export spreadsheets for a translator and reimport them a week later.

What separates Tolgee from a plain i18n library is in-context translation. Once the SDK is installed, you can hold the ALT or Option key, click any piece of text in your own app, and a dialog opens where you edit that string directly. The change syncs back to the platform. This works in development and, with the Tolgee Tools browser extension and an API key, in a deployed production environment too, which means a founder, a client or a non-technical colleague can fix wording without a developer, a pull request or a release.

The platform is available two ways. Tolgee Cloud is the hosted service with a free tier and paid plans, and the self-hosted version runs on your own servers from a Docker image. The core codebase is licensed under Apache 2.0, with a separate enterprise directory under Tolgee's own license covering advanced features like SSO and granular permissions.

Tolgee was founded in March 2021 by Jan Cizmar, a Czech developer who serves as CEO, alongside co-founder Marketa Chalupnikova. The company is based in Brno, Czechia. Cizmar has described the origin story consistently across interviews and conference talks: he started building Tolgee for his master's thesis in late 2019, after working at a company where localization was the single most common source of bugs because developers hated the process so much they left it until last.

The motivation shows in the product. Almost every design decision points at removing friction for the person writing the code rather than the person managing the translation project. Cizmar has spoken at FOSDEM and DevConf.CZ about the platform's internals, and the team develops in public on GitHub with an open roadmap, public issue tracker, and community Slack and Discord channels.

Tolgee is a small, independent company. It has not raised significant venture funding, and it has participated in Czech startup programs including CzechInvest's technology incubation scheme, which supported development of its AI translator. For some buyers, independence and a permissive core license are exactly the point, since it means no acquisition risk to a competitor and the guarantee that you can always take your data and run it yourself. For others, a team of this size is a genuine consideration when picking infrastructure for a decade.

The clearest fit is a small product team or technical founder shipping a web or mobile app into more than one language. If you already use React, Vue, Angular, Svelte, Next.js or a mobile framework, you install the SDK, wrap your strings, and get a working localization pipeline in an afternoon. The free tier covers 500 keys and three seats, which is enough to fully translate a marketing site, a small SaaS dashboard or an MVP before you pay anything.

Agencies and freelancers use it differently. Because in-context editing works in production, an agency can hand a client the browser extension and an API key and let them adjust their own copy after handover. That removes an entire class of low-value support tickets. Several public reviews from agency developers cite this specific workflow as the reason they chose Tolgee over alternatives.

It also suits teams with data residency or compliance constraints. The self-hosted build runs inside a corporate network or a regulated environment with translation data never leaving your infrastructure, and you can point machine translation at your own DeepL, Google or AWS API keys so text is not routed through Tolgee's account. Open-source projects are a fourth group worth calling out, since Tolgee offers its cloud service free to open-source projects and a free business plan for self-hosted open-source use.

In-context translation and one-click screenshots are the headline features. ALT-clicking a string opens an editor overlay; hitting the camera button captures a screenshot of your app with the target phrase highlighted and attaches it to the key. Context is the single biggest cause of bad software translation, and this solves it mechanically rather than by asking translators to guess. The classic example the team uses is the English word save, which maps to completely different words depending on whether it means storing a file, spending less money, or rescuing someone.

On the automation side, Tolgee supports machine translation through DeepL, Google Translate and AWS Translate, plus its own AI translator with brand context, where you supply a project description, tone notes and custom prompts so output matches your product voice. Translation memory suggests previously used translations with a similarity percentage, and auto-translation can fill new keys the moment they are created. Higher plans add an AI playground for testing prompt setups, glossaries for enforcing terminology, and automated quality assurance checks.

The developer tooling is the third pillar. The CLI extracts strings from source, pushes and pulls translations, and slots into CI. The REST API covers project setup and string management programmatically. Content delivery exports translations to a CDN or to your own AWS S3 or Azure blob storage so copy fixes reach users without shipping a new build. There is a Figma plugin for pulling real strings into designs, a Slack integration on higher tiers, webhooks, and an MCP server that lets AI coding assistants read and write your translation keys directly.

Tolgee Cloud is priced on three variables: keys, seats and machine translation credits. A key is one string identifier, counted once no matter how many languages you translate it into, which is a meaningfully friendlier model than the per-word or per-hosted-word pricing used elsewhere. Adding a tenth language costs you nothing in key terms. MT credits are roughly one credit per translated character for the standard providers, and are calculated from token usage for the AI translator.

The free plan gives 500 keys, three seats and 10,000 MT credits with no credit card, and it includes the core localization features, the SDKs, the Figma plugin and the MCP server. Team at 49 euros a month billed annually raises that to 2,000 keys and four seats and adds glossaries, webhooks, custom content storage and standard support. Business at 179 euros a month is the plan Tolgee flags as most popular, covering 5,000 keys, eight seats, one million MT credits, tasks, the AI playground, translation memory management and QA checks. Advanced at 499 euros a month covers 20,000 keys and twenty seats and unlocks branching, granular permissions, labels and the Slack integration. Enterprise is quote-based and is the only route to single sign-on, premium support, a dedicated Slack channel and an account manager.

Overages are pay as you go rather than a forced upgrade, which softens the plan jumps: 5 euros a month per extra 1,000 keys, 0.35 euros per extra 1,000 MT credits, and extra seats at 8, 14 or 22 euros a month depending on your plan. Annual billing gets two months free. Self-hosting the open-source build costs nothing in license fees and includes core localization plus machine translation via your own API keys, though advanced features like SSO and granular permissions require a paid license negotiated with the team. Budget honestly for the server, database, backups and patching before treating self-hosted as the free option.

The in-context workflow is the reason to pick Tolgee, and nothing else in this category does it as cleanly. Being able to click a string in your live app and rewrite it removes the slowest part of localization, which is the round trip between spotting a bad translation, finding its key, editing a file, and shipping a release. Reviewers repeatedly single this out, and one testimonial on Tolgee's own site describes handing the capability directly to clients so they can maintain their own copy.

Developer experience is the second real strength. The framework coverage is broad and the SDKs are properly native rather than thin wrappers, the CLI is well documented, and the platform assumes from the start that localization belongs in version control and CI rather than in an email thread. The MCP server being available on every plan, including free, is a forward-looking move that makes Tolgee unusually easy to drive from an AI coding agent.

Value is the third. A free tier that covers a genuine small project, a permissive core license, the option to self-host, free cloud access for open-source projects, and key-based pricing that does not punish you for adding languages together add up to a low-risk starting point. Reviewers on Capterra and G2 consistently describe it as fairly priced and note that the core functionality is solid rather than over-engineered, with an interface that stays minimal instead of burying the daily actions.

Tolgee is a developer tool wearing friendly clothing. The translator-facing side is genuinely approachable, but the setup is not: someone has to install an SDK, wrap strings, configure the CLI and wire content delivery. If your team has no developer, this is not a no-code product and you should look at a translation service rather than a translation management system.

Feature gating is the most common practical complaint. Tasks, which is how you assign and track translation work, sits on Business at 179 euros a month. Slack, labels and granular permissions sit on Advanced at 499 euros. Single sign-on is Enterprise only. For a five-person startup that just needs to assign translation work and restrict who can edit which language, that is a steep ladder. Third-party integration breadth is also thinner than the incumbents; reviewers have specifically noted the absence of the wider connector ecosystem that Crowdin and Phrase offer, and Tolgee has closed some of that gap over time but not all of it.

A few smaller things are worth flagging. Reviewers on G2 mention that some editor actions take more clicks than they should and that save times can feel slow on larger projects, and at least one Capterra reviewer reported placeholder escaping issues, which the vendor responded to publicly. Tolgee's review base is also small, at single-digit and low-double-digit counts on the major software directories, so there is less aggregate evidence to lean on than with a platform that has hundreds of reviews. Finally, if you rely on the SDK to fetch translations at runtime on the client, search engines may not see your translated content; use content delivery with server-side rendering or a build-time export if organic traffic in each locale matters to you.

Framework coverage is where Tolgee is strongest. The JavaScript SDK supports React, Vue, Angular, Svelte, Next.js, Nuxt and vanilla JavaScript, with an i18next plugin for teams already committed to that library. Native mobile SDKs cover iOS in Swift, Android in Kotlin including Compose Multiplatform, React Native and Flutter, with over-the-air translation updates so you can fix copy without an app store release. File format support covers JSON, PO, XLIFF, CSV, Apple strings and Android XML, so migrating an existing project in or out is a CLI command rather than a project.

Beyond code, the ecosystem is more selective. There is a Figma plugin for pulling live strings into design files, a Chrome extension for in-context editing on deployed sites, a Slack integration on Advanced and above, webhooks from Team upward, and an MCP server for AI assistants. Machine translation connects to DeepL, Google Translate and AWS Translate, and you can supply your own API keys, which is useful for cost control and for keeping data inside providers you already have agreements with.

What you do not get is a large app marketplace. Crowdin advertises hundreds of store apps and Phrase has an extensive partner network; Tolgee's approach is to give you a documented REST API, a CLI and webhooks and let you build what you need. For a technical team automating with GitHub Actions, Make or n8n, that is usually enough. For a localization manager who wants a prebuilt connector to an existing content or project management stack, it can mean building glue code that a competitor would have supplied.

Crowdin is the most direct commercial comparison. It launched in 2009, has a much larger ecosystem with hundreds of marketplace apps, strong document translation and community translation features, and a deep following in open source. It is closed source and cloud only, its pricing starts higher, and reviewers frequently mention a learning curve. Choose Crowdin when ecosystem breadth and vendor maturity outweigh the in-context editing and self-hosting that Tolgee offers.

Lokalise and Phrase sit at the enterprise end. Both are polished, both have serious workflow, quality assurance and vendor management tooling, and both cost considerably more once you scale. They suit organizations with a dedicated localization function and a professional translation supply chain. Locize is a closer philosophical match if you already use i18next, since it was built by the same team, and it emphasizes zero-operations managed delivery over self-hosting. SimpleLocalize is the budget-conscious option, starting around 12 euros a month with tasks, labels and Slack included on cheaper plans, though it does not match Tolgee's in-context editor.

On the fully open-source side, Weblate is the main alternative and is popular with open-source projects and organizations that want Git-native translation with a strong community. It is more oriented toward the traditional translation contributor workflow and less toward embedded in-app editing. If you want a free self-hosted platform and your workflow revolves around pull requests, look at Weblate. If it revolves around your running application, Tolgee is the better fit.