Tolgee

Self-hosted app localization, honestly reviewed. No marketing copy — just what you actually get when you run it yourself.

TL;DR

• What it is: Open-source localization platform — an alternative to Crowdin, Phrase, and Lokalise, with in-context editing and machine translation baked in [README][3].
• Who it’s for: Development teams building multilingual apps who are tired of managing .json or .po files by hand, and non-technical translators who need to work without touching code [README][3].
• Cost savings: Crowdin starts at $59/month; Phrase and Lokalise are in the same range and climb fast with seat and key counts. Tolgee self-hosted is free software on a VPS you already own [5][2].
• Key strength: In-context translation — translators (or clients) can click any text in the live app, ALT+click to open a translation dialog, and commit changes without touching a single file. No developer required for most translation updates [README][website].
• Key weakness: The self-hosted free tier is limited. Content Delivery (CDN) features, advanced AI translation, SSO, and anything past the baseline require either Tolgee Cloud or a negotiated licensed self-host plan. The vendor actively steers you toward their cloud [2].

---

What is Tolgee

Tolgee is a web-based localization platform built for software teams. The core problem it solves is the manual friction of app translation: developers extract strings into JSON files, export them, email them to translators, wait, receive back, reimport, rebuild, deploy. Every label change requires a developer task. Tolgee replaces that loop with a platform where translators work directly inside the running app [README][website].

The project positions itself as “an open-source alternative to Crowdin, Phrase, or Lokalise” — which are enterprise translation management systems costing $59–$250+/month for serious usage [README][5]. GitHub shows 3,855 stars for the platform repository, with a separate tolgee-js SDK repository that also has community traction.

The license situation needs a clear-eyed read. The badge on GitHub reads “Apache 2 / Tolgee EL” — a dual license. The core is Apache 2, which is genuinely permissive. But the “EL” (Enterprise License) tier covers advanced features, and the self-hosted pricing page [2] makes plain that some capabilities — including Content Delivery, certain AI features, and enterprise support — are either Tolgee Cloud-exclusive or require a paid licensed plan. This is the Tolgee business model, and it’s not hidden, but you need to read past the “open-source” headline to understand where the free tier ends.

The team is Prague-based, the founder and CEO Jan Cizmar is publicly active on the blog [1], and they’re hiring engineers — signs of a live company, not an abandoned repo.

---

Why people choose it over Crowdin, Phrase, and Lokalise

The 95 Capterra reviews average 4.6/5 with 96% positive sentiment [3][4][5]. The pattern across reviews is consistent: developers love it, translators and non-technical users are pleasantly surprised by the in-context editing, and the main complaints are about documentation gaps and occasional edge cases in SDK behavior.

The in-context editing argument. This is Tolgee’s strongest differentiator. Every other platform in the category works the same way: you manage strings in a web dashboard, export files, and rebuild. Tolgee’s SDK integrates into your JavaScript app so that ALT+clicking any visible string opens an inline translation dialog. Changes publish without a deployment. One reviewer on Capterra (Australia, health/fitness): “The devex for using translations in React is awesome, coupled with the ability for the Chrome extension to create translations is just great” [3]. This is the feature that gets cited most often when users explain why they switched.

One-click screenshots for context. The hardest problem in machine and human translation is context — a string that says “Close” could mean “shut the dialog” or “nearby” depending on the UI. Tolgee’s Chrome plugin takes a screenshot of the app with the string highlighted in one click, and that screenshot is attached to the translation key as context for the AI and human translators [README]. One developer on Capterra (US, computer software): “Love the ability to train the AI by providing overall app context, and specific key context” — then flagged that it “had occasional issues with improperly escaping placeholders” [3].

The cost argument against Crowdin and Phrase. Crowdin starts at $59/month [5]. Phrase and Lokalise are in the same ballpark and both charge by seat and by managed key volume — so the bill climbs as your team and translation catalog grows. Tolgee self-hosted is free software. For a startup adding three or four languages with a small team, the math is straightforward: $0 on Tolgee vs. $700+ per year on Crowdin, assuming you can handle Docker deployment [2].

The context accuracy argument against generic machine translation. The claim from the website — that the core reason for inaccurate translations is missing context — is not marketing noise. Machine translation of isolated strings performs worse than translation with screenshot context and key descriptions. Tolgee’s AI translator receives both, which is a meaningful technical advantage over uploading a flat JSON file to DeepL [README][website].

---

Features: what it actually does

In-context editing and Chrome plugin:

• ALT+click any string in your running app to open a translation dialog [README]
• Works in production environments — translators can edit the live app without code access [README]
• Chrome plugin generates screenshots in one click with highlighted strings [README]
• Screenshots attached to keys as AI and translator context [README]

Translation management:

• Web dashboard for managing all translation keys and languages
• Translation memory: suggests based on previously translated strings, includes similarity percentage, the key, and original text [README]
• Glossary support for consistent terminology across keys
• Import/export via CLI, REST API, or UI for .json, .po, Apple .strings, Android, Flutter [website]

Machine and AI translation:

• DeepL, Google Translate, and AWS Translate supported [README]
• Tolgee AI translator uses screenshot context for accuracy [website]
• Auto-translation on new keys via machine translation or translation memory [README]
• Bring your own API keys for ChatGPT, Claude, or other LLMs on self-hosted licensed plans [2]

Integrations:

• SDKs for React, Angular, Vue, Svelte, Next.js, vanilla JS [website]
• Figma plugin for designers to manage copy [website]
• Slack integration [website]
• REST API for CI/CD pipeline integration — one reviewer explicitly called out “the modern APIs I needed to fully integrate it into a CI/CD pipeline for deployment and version management” [3]
• Tolgee Tools Chrome extension for production translation [README]

Platform features:

• Team collaboration with translator roles
• Review/approval workflow (optional step between auto-translation and publish)
• Content Delivery / CDN feature for over-the-air translation updates — requires S3 or Azure Blob storage setup [1]

---

Pricing: SaaS vs self-hosted math

Tolgee Cloud:

• Free tier: available, no card required [website]
• Paid: usage-based from €0.00/month (specific tier pricing not published on the pricing page reviewed — contact or sign up to see current rates) [website]

Self-hosted:

• Community edition: free, Apache 2 licensed for core features [README]
• Licensed self-hosted: custom pricing, negotiated per team. Includes advanced features, AI translation via Tolgee Cloud credits (no own infrastructure for AI needed), and dedicated Slack support. Licensing based on keys and seats. Monthly or annual subscriptions [2].
• The vendor is explicit that self-hosting comes with maintenance overhead: updates, backups, incident response, and slower support are your problem [2].

Competitors for comparison:

• Crowdin: starts at $59/month [5]
• Phrase Localization Platform: €0.00 starting (per Capterra listing, but enterprise pricing applies for teams) [3]
• Lokalise: not in the source data — pricing not cited

The realistic math for a small dev team going multilingual:

Say you’re a 3-person startup adding 5 languages to a React app. On Crowdin, 3 seats at the team tier runs ~$200+/month once you have any meaningful key volume. On Tolgee Cloud, the free tier may cover early stages. Self-hosted, you pay $0 in licensing on a VPS you might already run (add $6–15/month if it’s dedicated) plus engineering time to deploy and maintain. Over a year: Crowdin ≈ $2,400+. Tolgee self-hosted ≈ $72–180 in VPS costs. That’s a meaningful difference for a bootstrapped team [2][5].

Caveat: if you need CDN/Content Delivery, you’ll also need an S3 bucket or Azure Blob — the setup is documented in a detailed tutorial [1] but adds infrastructure complexity.

---

Deployment reality check

The self-hosting guide [1] is detailed and usable. The CEO wrote it himself, which is either reassuring (founder-level care) or a flag about team size (founder has to write infra docs). The setup is Docker Compose plus Postgres, with optional S3/Azure for Content Delivery.

What you need:

• A Linux VPS with Docker and docker-compose
• Postgres (bundled in docker-compose or external)
• Optional: S3 bucket or Azure Blob for CDN features — requires AWS IAM setup [1]
• A domain and reverse proxy for HTTPS
• Optional: SMTP for email notifications

The S3 setup if you want Content Delivery is non-trivial for non-technical users. The tutorial [1] walks through creating an S3 bucket, setting bucket policies, creating IAM users, and wiring credentials through a config.yaml. This is a three-hour afternoon for someone who’s done it before; it’s a full day for someone who hasn’t. The tutorial exists and is current, but this is not a one-click deploy.

What can go sideways:

• The Chrome extension documentation is sparse. One Capterra reviewer described it as “difficult to figure out without hunting down the documentation page” [3][5].
• Placeholder escaping has been a reported bug — one reviewer flagged “occasional issues with improperly escaping placeholders” [3], which is exactly the kind of silent error that causes production bugs in translated apps.
• The vendor’s self-hosted pricing page [2] is unusually frank about steering you toward cloud: “Maintenance overhead. You are responsible for keeping the server running, applying updates, managing backups, and handling incidents.” They’re not wrong, but it reads as a soft sales page for their cloud tier dressed as documentation.
• Licensed self-hosted pricing is not published. You contact sales. This is fine for enterprise procurement; it’s opaque for a founder trying to budget.

Realistic time estimate: 1–2 hours for a developer who knows Docker to get a working instance. Add another 2–4 hours if you want Content Delivery via S3. Non-technical users need technical help for deployment.

---

Pros and cons

Pros

• In-context editing is genuinely unique. No other self-hostable localization tool gives translators and clients a live in-app editing experience with one-click deploy. This feature alone removes an entire category of developer interrupt [README][3].
• One-click screenshots with context. Machine translation accuracy improves materially when the model sees the UI element. The screenshot-to-key workflow solves the context problem in a way that flat-file exports can’t [README].
• Developer experience is well-reviewed. Multiple Capterra reviewers specifically call out React SDK integration and CI/CD API as clean and well-designed [3][4][5]. One Capterra reviewer: “fairly priced, highly capable and very reliable” [3].
• Translation memory reduces costs and improves consistency. Auto-suggestions with similarity scores, key reference, and original text make consistent terminology manageable at scale [README].
• Supports multiple MT providers. DeepL, Google Translate, AWS Translate — choose or combine, with your own API keys [README].
• Figma plugin. Designers can work with real copy in Figma without developer handoff for strings [website].
• 4.6/5 on Capterra across 95 reviews, 96% positive. Statistically credible positive signal [3][4][5].

Cons

• Free self-hosted tier is gated. CDN/Content Delivery, advanced AI translation, and enterprise features require either Tolgee Cloud or a custom licensed plan. You don’t know the price until you contact sales [2].
• Vendor actively steers toward cloud. The self-hosted pricing page reads as a reasons-to-use-cloud document. This isn’t deceptive — the costs of self-hosting are real — but the framing suggests the self-hosted path is second-class [2].
• Chrome extension documentation is sparse. Multiple reviewers flag this as a friction point [3][5]. For a tool whose key feature depends on the Chrome plugin working correctly, this is a notable gap.
• Placeholder escaping bugs reported. At least one Capterra reviewer hit production-impacting placeholder issues [3]. For i18n workflows, silent string errors are high-severity.
• Licensing opacity for licensed self-hosted plans. No public pricing. “Everything is negotiable” [2] is a red flag for teams that need predictable budgets.
• 3,855 GitHub stars is respectable but well below Crowdin’s ecosystem maturity. Fewer community integrations, fewer third-party guides, less Stack Overflow coverage for edge cases.
• Small company. Signs of a healthy company (founder-written docs, active blog, hiring) are also signs of a small one. Longevity bet is unproven compared to Crowdin or Phrase [1][README].

---

Who should use this / who shouldn’t

Use Tolgee if:

• You’re a developer team building a JavaScript/React/Vue app that needs localization and you want translators and clients to work without touching code or files.
• You’re paying $59–$200+/month for Crowdin or Phrase and your team is small enough that the self-hosted free tier covers you.
• You want in-context, production-environment translation editing — no other self-hostable tool in this category offers it cleanly.
• You have someone who can deploy Docker and optionally configure S3.
• You want AI translation with actual UI context, not just string-level machine translation.

Skip it (stay on Crowdin/Phrase) if:

• Your compliance team won’t approve self-hosted infrastructure or requires enterprise SLAs you can’t get from a contact-sales self-hosted plan.
• You need transparent, published pricing for budgeting before a sales call.
• You work primarily with non-JavaScript platforms where SDK depth matters (mobile-first, .NET, Python).

Skip it (use a simpler tool) if:

• You have one language to add and don’t need a platform — just i18next and a translator with a Google Sheet will do.
• You’re a non-technical founder who can’t deploy Docker and doesn’t have technical help. The setup is not one-click, and the documentation has gaps [3][5][1].

Skip it (use Weblate instead) if:

• You need a genuinely open-source, self-hostable localization platform with no enterprise gating and a track record of community governance. Weblate is the other serious open-source option in this space.

---

Alternatives worth considering

• Crowdin — the market leader for enterprise localization management. More mature, larger integration catalog, $59+/month, fully closed source. Better if you need proven enterprise support and don’t want self-hosting [5].
• Phrase Localization Platform — similar enterprise tier to Crowdin, strong TMS features, no self-hosting option [3].
• Lokalise — developer-friendly SaaS with good API, no self-hosting, pricing comparable to Crowdin.
• Weblate — genuinely open-source (GPL), self-hostable, strong community, no in-context editing, better fit for open-source projects that need community translation.
• Loco — simpler, cheaper SaaS for small teams; no in-context editing.
• Transloco + flat files — not a platform, but for small projects, the overhead of any platform may not be justified.

For a funded startup paying Crowdin bills and running a JavaScript app, the realistic comparison is Tolgee vs. Crowdin. Tolgee wins on in-context editing and cost; Crowdin wins on maturity, ecosystem, and predictable enterprise pricing.

---

Bottom line

Tolgee’s in-context translation is the genuine differentiator in this category. No other self-hostable localization tool lets a client or translator click text in the live app and commit a change without a developer. That feature alone is worth evaluating the platform for any JavaScript team that repeatedly fields “can we change this label?” tickets. The cost argument against Crowdin is real — the open-source core is free software on hardware you control. The catches are legitimate: CDN and advanced features require their cloud or an opaque licensed plan, the Chrome plugin documentation needs work, and at least one category of placeholder bug has reached production for users. For a developer-run team going multilingual on a budget, Tolgee is the strongest self-hostable option in the localization space right now. For a non-technical founder who can’t touch a terminal, it’s not the right tool without technical help — and that’s exactly the gap upready.dev exists to fill.

---

Sources

1. 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
2. Tolgee — “Self-Hosting Tolgee — Pricing & Plans”. https://tolgee.io/pricing/self-hosted
3. Capterra New Zealand — “Tolgee Pricing, Reviews & Features” (95 reviews, 4.6/5). https://www.capterra.co.nz/reviews/1039225/tolgee
4. Capterra Canada — “Tolgee Pricing, Reviews & Features” (95 reviews, 4.6/5). https://www.capterra.ca/reviews/1039225/tolgee
5. 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-hosted pricing: https://tolgee.io/pricing/self-hosted