Tolgee
Tolgee offers in-context translating, translation memory, machine translations as a self-hosted developer tools.
Self-hosted i18n, honestly reviewed. What you actually get when you run Tolgee on your own server instead of paying Crowdin or Phrase.
TL;DR
- What it is: Open-source localization platform — think Crowdin or Phrase, but self-hostable, with a developer SDK that lets translators edit strings directly in the running app without touching code [README].
- Who it’s for: Dev teams shipping multilingual apps who are tired of the edit-export-upload cycle with expensive SaaS platforms. Also useful for non-technical clients who need to manage translations without dev help [3].
- Cost savings: Crowdin starts at $59/mo; Phrase is enterprise-tier pricing. Tolgee self-hosted runs free on a VPS you already own. Cloud plan starts at €0 with usage-based upgrades [2][5].
- Key strength: In-context editing — translators (or clients) hold ALT, click any string in the live app, and edit it on the spot. No JSON files, no spreadsheets, no back-and-forth with developers [README].
- Key weakness: The self-hosted free tier has real limitations (single CDN configuration, no SSO, enterprise features require a negotiated license). Support on self-hosted is explicitly slower than cloud because the team can’t access your servers [2].
What is Tolgee
Tolgee is a web-based localization platform built for development teams. You add a JavaScript SDK to your app, connect it to a Tolgee instance, and from that point forward anyone — developers, translators, or clients — can edit translations directly inside the running application without touching a file or triggering a deployment.
The README describes it as “an open-source alternative to Crowdin, Phrase, or Lokalise” [README]. That’s the right frame. Those three tools dominate the commercial market and charge per seat or per feature at rates that scale painfully once you’re managing multiple projects and languages.
What makes Tolgee worth evaluating is three things. First, the in-context editing workflow: translators hold ALT and click any text in the live UI to get an edit dialog, with a one-click screenshot button to generate context automatically [README]. This is the workflow gap that breaks every other tool — context is why translations are inaccurate, and every other platform tries to solve it with screenshot uploads, Figma links, or comments. Tolgee solves it by making the running app the editor. Second, the SDK-first integration model: Tolgee embeds into React, Angular, Vue, Svelte, Next.js, Flutter, Android, and vanilla JS — it’s not just a file sync service, it’s wired into your component tree [website]. Third, the machine translation pipeline: DeepL, Google Translate, AWS Translate, and OpenAI/Claude-compatible LLMs are all supported, and Tolgee uses screenshot context to improve MT accuracy rather than just running a generic translation [README].
The platform sits at 3,855 GitHub stars and claims 40,000+ users worldwide [merged profile][website]. Capterra shows 95 verified reviews with a 4.6/5 average and 96% positive sentiment [3][4][5].
Why people choose it
The Capterra reviews [3][4][5] tell a consistent story about the two main purchase drivers.
The developer experience is unusually good. A software engineer (Australia, health/wellness): “The devex for using translations in react are awesome, coupled with the ability for the chrome extension to create translations is just great” [3]. A developer (US, computer software): “I very much enjoyed the platform. It is fairly priced, highly capable and very reliable” [3]. The React SDK wrapper is what gets cited most often — it’s genuinely lower friction than competitors.
The in-context editing is the differentiator. The Capterra overview states the core thesis plainly: “The core reason behind all inaccurate translations is missing context” [3]. Tolgee’s bet is that if you give translators a direct editing interface inside the live app with automatic screenshot generation, you eliminate most of the back-and-forth. The homepage testimonials include: “I really like the in-context editing feature of Tolgee. This way our clients have an easy to use way to edit translations of the platforms we build” from a Project Leader [website]. A production developer at Curipod (an education platform) cited a concrete result: “We were seeing hundreds of React crashes per day from browser translation. Tolgee brought that number to zero overnight” [website].
Versus Crowdin. Crowdin is the category leader with a richer feature set and a larger ecosystem, but it starts at $59/mo and scales up from there [5]. Tolgee’s self-hosted path costs your server time and nothing else. For a team managing 3–5 projects with modest key counts, Crowdin’s pricing is genuinely painful.
Versus Phrase and Lokalise. Both are enterprise-focused, both are expensive. Phrase in particular is oriented around large localization teams with complex workflows. Tolgee doesn’t match them on feature depth, but it’s targeting a different customer: the dev team that wants to ship in five languages without a dedicated localization budget.
Versus doing it manually. The website’s most pointed argument is the simplest: every translation request without Tolgee becomes a developer task. Someone finds a typo in the German version, files a ticket, developer edits a JSON file, deploys, done. Tolgee lets the person who found the typo fix it in 30 seconds from any browser [website].
Features
Core translation workflow:
- Web-based translation editor with key management, search, filter by state [README]
- In-context editing via ALT+click in the live app — works in production with the Chrome plugin [README]
- One-click screenshot generation tied to specific translation keys [README]
- Translation memory with similarity percentage and source text shown in suggestions [README]
- Auto-translation on new keys using memory or machine translation [README]
- Comment threads and review states per key [README]
Machine translation:
- DeepL, Google Translate, and AWS Translate built-in [README]
- OpenAI, Claude (via bring-your-own API key), and other LLMs for AI translation [2]
- Context from screenshots improves MT accuracy rather than raw text-only translation [README]
- AI translator can be routed through Tolgee Cloud even on self-hosted (credits-based) [2]
Developer integrations:
- SDKs: React, Angular, Vue, Svelte, Next.js, vanilla JS, Flutter, Android, Apple strings, .po, PHP [website]
- Figma plugin for design-to-translation workflow [website]
- REST API for CI/CD integration [website]
- CLI for file import/export [website]
- Slack integration for notifications [website]
Self-hosting:
- Docker Compose deployment with Postgres [1]
- Kubernetes support [2]
- S3 (AWS) or Azure Blob for Content Delivery (CDN) configuration [1]
- Single CDN configuration on free self-hosted tier [1]
- License key activation for enterprise features on self-hosted [2]
What’s gated:
- SSO is not mentioned in the free tier — enterprise features require a negotiated licensed plan [2]
- Multiple CDN configurations require paid licensed self-hosting [1]
- Support on self-hosted is more limited than cloud by design [2]
Pricing: SaaS vs self-hosted math
Tolgee Cloud:
- Free tier: available, no card required [website]
- Usage-based pricing: €0 starting price per Capterra [3][4][5]
- MT credits: usage-based, purchased separately — even self-hosted instances can route through Tolgee Cloud for AI translation [2]
- Enterprise/custom: contact sales
Self-hosted (free):
- Software: open-source, Apache 2 / Tolgee EL dual license [README]
- VPS cost: whatever you’re already paying, or $5–15/mo on Hetzner/Contabo
- CDN: one S3/Azure Blob configuration included [1]
- Enterprise features: negotiated licensed plan, custom pricing, monthly or annual [2]
Crowdin for comparison:
- Starter: $59/mo (up to 10 projects, 10 seats)
- Team: $119/mo
- Business: $229/mo+
- Enterprise: custom
Phrase for comparison: Enterprise pricing, not publicly listed, consistently described in reviews as the expensive end of the market.
Concrete savings for a small dev team:
A 5-person dev shop managing 3 multilingual apps on Crowdin Starter pays $59/mo ($708/year). On Tolgee self-hosted: $0 in licensing plus whatever the VPS costs. If it’s on existing infrastructure: $0. Even if you spin up a dedicated $6/mo VPS: $72/year versus $708. That’s $636/year saved, and you own the data [2].
The catch: licensed self-hosting for advanced features (if you need them) is custom-priced and requires a conversation with Tolgee. There’s no public tier chart for enterprise self-hosting, which makes it harder to compare directly [2]. The free self-hosted version is real and functional, but the CDN limitation (one configuration) will matter if you’re serving translations from edge locations.
Deployment reality check
Tolgee publishes a working Docker Compose example at https://github.com/tolgee/docker-compose-content-delivery-example [1]. The stack is straightforward: a Java/Kotlin backend, Postgres for data, and optionally S3 or Azure Blob for content delivery.
What you actually need:
- A Linux VPS with Docker and docker-compose (the backend is JVM-based, so budget at least 1–2GB RAM; 2–4GB recommended)
- Postgres (bundled in the docker-compose example or external)
- A domain and reverse proxy (nginx or Caddy) for HTTPS
- For CDN features: an AWS S3 bucket or Azure Blob container with appropriate IAM policies [1]
The S3 setup adds steps. The official blog guide [1] walks through creating an S3 bucket, setting a public bucket policy for CDN access, creating an IAM user with scoped permissions, and wiring the credentials into config.yaml. It’s not complicated if you’ve used AWS before, but it’s five or six steps that a non-technical user will struggle with. The guide exists and is clear, which is better than average for self-hosted tools.
Config is file-based, not environment variables. The recommended pattern is mounting a config.yaml into the container rather than setting environment variables [1]. This is actually a reasonable security choice (env vars can appear in process listings), but it’s a different pattern than most Docker-deployed tools and worth knowing upfront.
What can go sideways:
- The Chrome extension requires separate installation and setup — the documentation exists but isn’t immediately obvious [3][4][5]. One Capterra reviewer called it out explicitly: “The chrome extension was difficult to figure out without hunting down the documentation page” [3].
- Placeholder escaping has occasional bugs. A Capterra reviewer (April 2025) noted “occasional issues with improperly escaping placeholders” [3]. This is a real edge case for ICU message format users.
- Tolgee’s own pricing page actively discourages self-hosting for most teams: “We recommend Tolgee Cloud for most teams. It is faster to get started, requires no maintenance, and gives you the best support experience” [2]. That’s an honest admission that self-hosted support is slower — the team can’t access your servers to diagnose problems.
Realistic time estimate for a technical user: 45–90 minutes to a working instance. Add another 30–45 minutes if you’re wiring up S3 for CDN. For a non-technical founder: budget an afternoon, or have a developer handle the initial setup.
Pros and Cons
Pros
- In-context editing is genuinely different. The ALT+click workflow in the live app is the cleanest solution to the translation context problem in the category. Competitors require screenshot uploads; Tolgee generates them automatically and ties them to specific keys [README].
- 40,000+ users, 4.6/5 on Capterra. 96% positive sentiment across 95 verified reviews is a reliable signal [3][4][5]. The product works in production.
- SDK-first, not file-sync. Tolgee embeds into your component tree. This isn’t just another platform that reads your
en.jsonfile — it’s wired into React, Next, Angular, Flutter [website]. - MT with context. Using screenshots to inform machine translation is a meaningful improvement over raw string translation. This matters most for UI strings where a word like “Set” is ambiguous without visual context [README].
- Self-hosted free tier is real. The free open-source version gives you the full translation workflow, MT integration, in-context editing, and SDK access. The limitations (single CDN config, no enterprise SSO) matter at scale, not for a typical small team [1][2].
- Figma integration and CLI. The combination of Figma plugin for designers and CLI for CI/CD covers the two most common integration points in modern web development [website].
- Translation memory built-in. Consistency across projects and keys without extra tooling [README].
Cons
- Chrome extension has a UX problem. Multiple reviewers flag that the extension isn’t self-explanatory — you need to hunt for the documentation to understand the setup [3][4][5]. For a tool that’s supposed to lower the barrier for non-developers, this is a meaningful gap.
- Placeholder escaping bugs. At least one Capterra reviewer hit placeholder issues in April 2025 [3]. ICU message format is complex and error-prone at the edges.
- Enterprise self-hosting is opaque. There’s no public pricing chart for licensed self-hosted plans. Everything above the free tier is “contact us” [2]. This makes budget planning harder and sales-cycle-dependent.
- Self-hosted support is explicitly limited. Tolgee’s own documentation says the team can’t diagnose self-hosted issues remotely [2]. If something breaks in production at 2am, you’re on your own unless you’re on cloud.
- Single CDN config on free self-hosted. The free tier locks you to one S3/Azure Blob configuration. Multi-region or multi-bucket setups require a licensed plan [1][2].
- No public enterprise pricing. Unlike Crowdin or Phrase, which at least show tier names, Tolgee’s licensed self-hosted plans are fully custom/negotiated [2]. “Everything is negotiable” is flexible but not efficient for teams that need a quick budget number.
- 3,855 GitHub stars is modest. For context, Crowdin has been around longer and is a commercial product, so stars aren’t a direct comparison. But Tolgee’s GitHub presence is smaller than category leaders. The product is real but the community is still growing [merged profile].
Who should use this / who shouldn’t
Use Tolgee if:
- You’re a dev team shipping a multilingual web or mobile app and you’re tired of the JSON-file-email-spreadsheet translation workflow.
- Your clients or non-technical teammates need to manage translations without filing tickets — the in-context editing workflow is genuinely usable by non-developers [3].
- You want to escape Crowdin’s $59+/mo pricing and self-host on infrastructure you already have.
- You’re building in React, Next.js, or Flutter — the SDKs for these are the most mature [website].
- You’re comfortable running Docker Compose and can handle an initial setup of 60–90 minutes.
Consider staying on Tolgee Cloud (not self-hosted) if:
- You want the in-context editing and SDK features but don’t want to manage infrastructure or deal with slower support.
- Your team is small and the cloud free tier covers your key count.
Skip it (use Crowdin) if:
- You need a mature, enterprise-supported localization workflow with a large translator marketplace, deep CMS integrations, and a proven support track record.
- Your localization team is large enough that the per-seat pricing difference stops mattering against support and feature maturity.
Skip it (use Phrase or Lokalise) if:
- You’re at the scale where translation is a first-class business function with dedicated localization managers and complex review workflows.
Skip it entirely if:
- You’re a non-technical founder with no developer to handle deployment and integration — Tolgee’s self-hosted value requires someone who can write
docker-compose upand wire an SDK into your codebase. - You need guaranteed SLA-backed support on self-hosted infrastructure. Tolgee is explicit that this isn’t available [2].
Alternatives worth considering
- Crowdin — the category standard. More integrations, larger translator community, mature enterprise workflows. Starts at $59/mo and goes up. Closed source, no self-hosting option [5].
- Phrase — enterprise-tier platform with deep CAT tool integration and translation memory across very large key counts. Expensive, not self-hostable, oriented toward dedicated localization teams.
- Lokalise — similar positioning to Phrase, strong CI/CD integrations, per-seat pricing. Also closed source.
- Weblate — the other major open-source localization platform. More translator-focused than developer-focused. GPL-licensed, self-hostable, strong for community/open-source projects. Lacks Tolgee’s in-context SDK approach.
- i18nexus — cloud-only, developer-friendly, integrates tightly with i18next. Simpler than Tolgee but no self-host option.
- Loco (localise.biz) — simpler SaaS option for smaller projects. Cheaper than Crowdin but also less capable.
For a development team choosing between open-source self-hosted options, the realistic comparison is Tolgee vs Weblate. Tolgee wins if your priority is the in-context editing SDK and developer workflow. Weblate wins if you’re managing a large community translation effort with many external contributors and want a more translator-centric interface.
Bottom line
Tolgee solves a real problem — the context gap that breaks most localization workflows — with a genuinely different technical approach. Instead of trying to add context via screenshot uploads or comment fields in a web UI, it embeds into your running application and makes the app itself the translation editor. That’s not a marginal improvement; it’s a workflow change that lets non-developers edit production translations without touching code. The Capterra reviews back it up: 96% positive, with the in-context editing and React SDK cited consistently as the reasons people chose it.
The trade-offs are real: the self-hosted free tier has CDN limitations, enterprise pricing is opaque, the Chrome extension needs better onboarding, and self-hosted support is explicitly weaker than cloud. If you’re a solo developer or small team spending $60–$150/mo on Crowdin or Phrase, the math for switching is obvious. If you need enterprise-grade localization workflows with SLA-backed support on self-hosted infrastructure, Tolgee can offer that — but you’ll be negotiating a custom license, not picking a plan from a pricing page.
Sources
- Jan Cizmar (Tolgee Blog) — “Ultimate guide to self-hosting Tolgee with Content Delivery (CDN)” (Feb 17, 2025). https://tolgee.io/blog/2025-02-17-self-hosting-tolgee-with-content-delivery
- Tolgee — “Self-hosted — Tolgee Pricing”. https://tolgee.io/pricing/self-hosted
- Capterra New Zealand — “Tolgee Pricing, Reviews & Features” (95 reviews, 4.6/5). https://www.capterra.co.nz/reviews/1039225/tolgee
- Capterra Canada — “Tolgee Pricing, Reviews & Features” (95 reviews, 4.6/5). https://www.capterra.ca/reviews/1039225/tolgee
- Capterra UK — “Tolgee Pricing, Cost & Reviews” (95 reviews, 4.6/5). https://www.capterra.co.uk/software/1039225/tolgee
Primary sources:
- GitHub repository: https://github.com/tolgee/tolgee-platform (3,855 stars, Apache 2 / Tolgee EL license)
- Official website: https://tolgee.io
- Self-hosting documentation: https://docs.tolgee.io
Features
Integrations & APIs
- Plugin / Extension System
Category
Related Developer Tools Tools
View all 181 →Neovim
97KThe hyperextensible Vim fork that rewards the time you invest — sub-100ms startup, modal editing, total customization, and no licensing fees.
Hoppscotch Community Edition
78KOpen-source API development ecosystem — lightweight, fast alternative to Postman with REST, GraphQL, WebSocket, and real-time API testing.
code-server
77KRun VS Code on any machine and access it through a browser — code from your iPad, Chromebook, or any device with a web browser.
Appwrite
55KOpen-source backend-as-a-service with authentication, databases, storage, functions, and messaging. Self-hosted Firebase alternative for web and mobile apps.
Gitea
54KLightweight, self-hosted Git service with code hosting, pull requests, CI/CD, package registry, and project management. GitHub alternative that runs on a Raspberry Pi.
Gogs
48KA painless, lightweight, self-hosted Git service written in Go. Minimal resource usage, easy setup, and runs on anything from a Raspberry Pi to a VPS.