I run a web agency. We build sites on Vercel, mostly SvelteKit and Next.js. And for the last two years, the same conversation keeps happening with clients.
"Why is our Weglot bill €300 a month?"
Fair question. Weglot works. I'll give them that. But at some point you look at the invoice and realize you're paying more for translation than for hosting, analytics, and email combined. And the pricing only goes up from there. Add a few languages, grow past their word limits, and suddenly you're on a plan that costs more than some people's rent.
I kept telling clients it was the best option. Because it was. The alternatives were either janky WordPress plugins, expensive TMS platforms that need a dedicated localization engineer, or hand-rolling i18n with JSON files and crying into your keyboard at 2am.
None of that worked for the kind of sites we build.
The Vercel problem
Here's the other thing that bothered me. We deploy on Vercel. Edge functions, serverless, the whole modern stack. And the translation tools that exist either don't work with that architecture, or they want enterprise pricing to make it happen.
Weglot works by injecting a client-side script. Fine for basic sites, but it flickers on load, it doesn't translate server-rendered content properly, and search engines see the original language first. Your translated pages are basically invisible to Google until the JavaScript runs.
The "proper" way to do multilingual on Vercel is either Vercel's own i18n routing (which just handles URL prefixes, not actual translation) or you pipe everything through a translation management system and maintain parallel content. That's a part-time job nobody wants.
I wanted something that sits in front of your site as a reverse proxy. Visitor hits /fr/pricing, the proxy grabs your English page, translates it server-side with AI, caches the result, and serves it with correct hreflang tags and translated meta descriptions. The original site doesn't change at all. No build step, no framework plugin, no content duplication.
So I built it.
What LangRelay actually does
LangRelay is a Cloudflare Worker that acts as a reverse proxy for your site. When someone visits a translated URL:
- The worker fetches your original page
- AI translates the text content (not the code, not the markup, just the human-readable text)
- It adds proper SEO tags — hreflang, localized canonical URLs, translated meta descriptions
- The translated page gets cached at the edge
- Subsequent visitors get the cached version in under 50ms
Your original site is never touched. You don't install anything. You point a DNS record or add a script tag, pick your languages in the dashboard, and you're done.
The part I'm most stubborn about: you only pay for unique translations. If the same sentence appears on your homepage, your about page, and your footer, it gets translated once. Cached hits are free. Model upgrades are free. If I improve the AI and re-translate your site with a better model, you don't pay again.
Weglot charges by total words across all pages. So that footer with "Copyright 2026 Acme Inc" gets counted on every single page. Multiply that by 12 languages and you're paying to translate the same five words hundreds of times. That always felt wrong to me.
Building in stealth
I've been working on this for months. Didn't announce anything, didn't launch on Product Hunt, didn't tweet about it. Just kept building and testing.
Right now LangRelay is live on a handful of client sites. Real production traffic, real users, real translated pages being served. The results have been good. Pages load fast because everything is cached at the edge. Google is indexing the translated versions. Clients are happy because their translation costs dropped significantly.
One site that was paying around €200/month on Weglot is now on LangRelay for a fraction of that. Same languages, same coverage, better SEO because the translations are server-rendered from the start.
I'm not going to pretend the product is finished. There are rough edges. The dashboard needs work. The glossary feature is basic. But the core — the proxy, the translation engine, the caching layer, the SEO — that part works well enough that I'm comfortable putting real client sites on it.
What's next
I'm opening this up. Not with a big launch, just gradually. If you're an agency owner or a developer who's been paying too much for translation, or you've been avoiding multilingual entirely because the tooling is bad, I'd like you to try it.
There's a free tier. 10 pages, 1 language, 10,000 words per month. Enough to test it on a small site and see if it works for you.
Paid plans start at €19/month with 100K words included. If you go over, there's a per-word overage rate — no surprise tier upgrades, no "contact sales" gates.
I'm building this because I needed it. My clients needed it. And I think a lot of other people need it too. Website translation shouldn't cost a fortune, and it shouldn't require a dedicated engineering effort. Add a DNS record, pick your languages, move on with your life.
That's the whole idea.