The cheapest and most efficient i18n for agentic coding
Keyless i18n your coding agent gets right the first time: one file to edit, no locale JSON to load, no merge conflict, from €4 a month.
5-Minute Setup
Your whole i18n implementation in 5 minutes. Not kidding.
No Keys System
Your brain shouldn't be involved for such meaningless tasks.
Global From Day One
Ship to international markets without extra effort.
Setup i18n, keep your codebase clean and ship fast
Multiple PITA exist with the current i18n solutions. Pains are so big that you delay i18n as much as possible.
It consumes time you should spend on the beach
Setup the keys' system: at least 1 hour of senior dev time
Back and forth for each new key: 1 minute per key, x500 keys, x2 languages = 17 hours
With i18n-keyless, you don't care about the i18n system at all. You can go on the beach and order a mojito.
Today most of the systems use keys to translate the text
This is painful to generate. This is painful to maintain.
When you see a text in the app, and you want to update it, you need to find the corresponding key.
With i18n-keyless, you don't care about the i18n system at all. Use your brain for something more interesting.
Leave translations to AI
With the key system, you need to manage the translations in the app.
Otherwise your users will see someheader.welcome.title in their app.
With i18n-keyless, worst case scenario offline you see untranslated real words. Normal pace, you don't care about the i18n system at all. Just write good headlines, focus on marketing, get sales.
Codebases are beautiful... until i18n keys' infection
With the key system, when you read the code and the content, you have to read keys, not natural language.
Sometimes you should make a fix because a sentence is not grammatically correct. But you don't know that because you read keys.
With i18n-keyless, you don't care about the i18n system at all. Codebase stays beautiful.
What if the service shuts down?
Nothing you have goes away.
SDKs are MIT
Every SDK stays on npm, PyPI, RubyGems, pub.dev and the other registries. Fork them, keep them, they are yours.
Server is open source
The very same Docker image that runs i18n-keyless.com. Run it on your machine with SQLite or Postgres, your AI key, your data at home.
One project free forever
€30 once for unlimited projects. The licence is verified offline — the server never calls home. Export all translations to JSON at any time.
MCP server to cover all your i18n needs
Plus six other reasons why the system is so optimized for agentic coding.
No name to invent
A key is a naming decision with no right answer. Two agents write login.submit and auth.button.send for the same button. Without keys there is no decision to make.
One file to edit, not twenty
A keyed change touches the component plus every locale JSON, and agents forget files. Here a copy change is a single-file edit.
Zero merge conflicts in agent fleets
Five agents on five branches all edit en.json: a guaranteed conflict. There is no shared file, so worktrees and background agents never collide.
Cheap in tokens
To reuse a key, an agent must load the locale files into context on every edit. Keyless loads nothing: the text is right there in the component.
Readable diffs
The pull request shows real sentences. A human or a review agent can judge the copy without digging for what a key stands for.
One rule in your CLAUDE.md
Every i18n rule you write is a rule the agent can break. This one needs a single line — and the fallback is the real text, so a miss cannot show a.b.c on screen.
The only i18n rule your agent needs
# CLAUDE.md Write the UI copy in English and wrap it in <T>…</T> from i18n-keyless-react. That is the whole i18n setup.
Your agent can operate it, not only write it
Through the MCP server, the agent lists missing translations, reads one, overrides a bad one, or imports the copy you already have — with its own OAuth login, no key to paste.
Read the MCP guideLaunch globally from day one
Stop delaying i18n because of complexity. Implement in 5 minutes, not days. Maintain your development velocity as if i18n wasn't implemented.
- 5-minute implementation. Not kidding.
One simple config object and you're done.
- Maintain development velocity
Code as if i18n wasn't implemented at all.
- Ship to global markets
No refactoring needed as you scale to new languages.
- SSR & SEO ready
Works with Next.js, Remix, TanStack Start and Astro. Server-render any language for search engines — same 5-minute setup.
Traditional i18n
i18n-keyless
It's no brainer to setup
Just the simplest config object and you're ready to go global.
And even more, your agent will set it up for you.
1. Basic Setup
import * as I18nKeyless from "i18n-keyless-react";
I18nKeyless.init({
API_KEY: "YOUR_API_KEY",
storage: window.localStorage,
languages: {
primary: "en",
supported: ["en", "fr"],
},
});2. Component Usage (React here; the same in Vue, Angular, Flutter, Swift and Kotlin)
import { I18nKeylessText } from "i18n-keyless-react";
// import { T } from "i18n-keyless-react"; // also works
// Just write in your primary language
<I18nKeylessText>
Welcome to our website
</I18nKeylessText>3. Server-side Usage (Node.js here; the same call in Laravel, Rails, Python and Go)
import { awaitForTranslationOrFallbackToOriginal, type Lang } from 'i18n-keyless-node';
const title = await awaitForTranslationOrFallbackToOriginal('Come see my app!', user.lang as Lang);
const body = await awaitForTranslationOrFallbackToOriginal('So many new stuff!', user.lang as Lang);
await sendExpoPushNotification({ user, title, body });4. For text outside components
import { getTranslation } from "i18n-keyless-react";
export default function Navigation() {
return (
<HomeTabs.Navigator>
<HomeTabs.Screen
options={{ tabBarLabel: getTranslation("Welcome") }}
name="WELCOME"
/>
</HomeTabs.Navigator>
);
}5. Pick your SDK: one install line, the same API everywhere
- React, React Native, Expo, Next.js, Remix, TanStack Start, Astronpm install i18n-keyless-react
- Vue, Nuxtnpm install i18n-keyless-vue
- Angularnpm install i18n-keyless-angular
- Plain HTML, Svelte, Alpine, htmx, one script tagnpm install i18n-keyless-browser
- Node.jsnpm install i18n-keyless-node
- Laravel, PHPcomposer require i18n-keyless/laravel
- Rails, Rubybundle add i18n-keyless-rails
- Flutter, Dartflutter pub add i18n_keyless
- Swift, iOS, macOS, VaporSwiftPM: arnaudambro/i18n-keyless-swift
- Kotlin, Android, Ktorio.github.arnaudambro:i18n-keyless-kotlin
- Python, Django, Flask, FastAPIpip install i18n-keyless
- Gogo get github.com/arnaudambro/i18n-keyless/ports/go/v3
How translations work
i18n-keyless is a library, combined with an API service, that allows you to translate your text without the need to use keys.
If the translation is not found, there will be an asynchronous fetch to I18nKeyless' API to get the translation by an AI API call.
This operation is made usually in dev mode, while you're coding. If you missed one, in production only the first user missing the translation will see a glitch, then it'll be available instantly to users all over the globe.
Your store listing in every language, on every release
Every update means new release notes, a description and keywords — for each language, on both stores. That is what stops you from adding languages. Not the app.
- Release notes for every language, in one call
Write the update once. awaitForTranslation returns it in each language you sell in, and your store config feeds it to EAS or fastlane.
- 48 languages, mapped to the App Store slots
pt-BR and pt-PT are two listings; en-GB, en-AU and en-CA are three. toAppStoreLocale puts each translation in the right one.
- Fix a bad translation without a release
In the app, one override in the dashboard is live at once — no hotfix build, no store review. In the listing, re-run the push.
- Not priced per word
A translator charges about €0.10 a word. 2,000 strings times 10 languages is a budget a solo developer does not have. Here words are not counted.
The whole recipe for an Expo app
// store.config.js — EAS Metadata awaits an async config
const { init, awaitForTranslation, toAppStoreLocale } = require("i18n-keyless-node");
const LANGS = ["en", "fr", "de", "ja", "pt-BR"];
const DESCRIPTION = "Bibi translates your app while you sleep…";
module.exports = async () => {
init({ API_KEY: process.env.I18N_KEYLESS_KEY, languages: { primary: "en", supported: LANGS } });
const info = {};
for (const lang of LANGS) {
info[toAppStoreLocale(lang)] = {
title: await awaitForTranslation("Bibi", lang, { context: "App Store app name, 30 characters max" }),
subtitle: await awaitForTranslation("Keyless i18n", lang, { context: "App Store subtitle, 30 characters max" }),
description: await awaitForTranslation(DESCRIPTION, lang, { context: "App Store description" }),
};
}
return { configVersion: 0, apple: { info } };
};The same call writes fastlane metadata files for Google Play. Caps are advisory: put the limit in the context and check the length before you push.
Interactive Sandbox
See how i18n-keyless works in real-time
Open a live, editable StackBlitz project to see translations resolve as you type.
Why I built this
"I always postponed i18n until my client wanted it so much that I had no choice. I hate reading the code with i18n keys inside, and I hate maintaining JSON files as much as seeing how expensive are i18n translations providers. So I had to build mine, and I FREAKING LOVE IT. All my apps are i18n by default now."
Arnaud Ambroselli
Founder of i18n-keyless
Simple, predictable pricing
One flat price for your scale.
Every feature in every plan.
No per-word fees, no token math.
Up to 1,000 monthly active users
For indie apps and new products
Up to 5,000 monthly active users
For apps finding their audience
Up to 50,000 monthly active users
For apps with real traction
Up to 500,000 monthly active users
For products at scale
Or own it: free to install, €30 once for unlimited projects.
The same API, dashboard and MCP server, in one Docker container with SQLite, on your server, with your own Mistral key. One project is free forever; a €30 lifetime licence unlocks unlimited projects. No subscription, no usage cap, updates for life.
Every plan includes:
All 48 languages. The whole world, from day one
Unlimited translations, AI included. New strings translate themselves in production
Live translation of UGC (user-generated content)
Upgrade, downgrade or cancel anytime
More than 500,000 monthly active users? Contact us
What the same app costs somewhere else
Move the two sliders. The bill is for your app in all 48 languages.
| Solution | Your monthly bill | You pay per | Translation | Languages |
|---|---|---|---|---|
| i18n-keylessus | €4/ month | Monthly active users Win win. | AI, unlimited, included | All 48, from day one |
| Crowdin | €88/ month1 | Hosted words: your words × 48 | Crowdin AI, inside a quota | You add each one |
| Intlayer | €27/ month2 | Seats and projects | AI generation, inside a quota | You add each one |
| Lingo.dev | €91/ month3 | A base fee, then AI tokens | The AI bill goes straight to you | You add each one |
| Locize | €77/ month4 | Words × 48, and CDN downloads | You bring your own AI key | 10 on the largest published plan |
| Lokalise | €132/ month5 | Seats, then keys | Translators billed apart | You add each one |
| Phrase | €483/ month6 | Seats, then managed words | From $0.06 a word, billed apart | You add each one |
| Smartling | €552/ month7 | Every word, every language | AI translation from $0.06 a word | You add each one |
| Tolgee | €49/ month8 | Keys and seats | Machine translation credits | You fill each one |
| Weglot | €1,678/ month9 | Words × languages, on a web page | Machine, inside a quota | 20 on the largest published plan |
Public list prices, read in August 2026. Vendors change them, so please check. Each bill is our reading of the published ladder, and we always take the cheapest paid plan that holds your app. When a vendor bills the first translation once, we spread it over twelve months.
1 Crowdin publishes €44 a month for 60,000 hosted words on the Pro plan, billed annually. A hosted word is one word in one target language. We repeat that block price above 60,000.
2 Intlayer is open source. The $29 a month buys the CMS and the AI generation quota. Above the quota you supply your own AI key, so we add the model bill at $10 for each million tokens, spread over twelve months.
3 Lingo.dev publishes $99 a month plus $2 for each million tokens of infrastructure, and it passes the model bill to you. We count 1.4 tokens for each word and $10 for each million tokens of model cost. The base fee is most of this number.
4 Locize publishes €7, €19 and €49 a month for 15,000, 30,000 and 60,000 words, with 1, 3 and 5 million CDN downloads. Above that it publishes $0.50 for each 1,000 extra words and $5 for each million extra downloads. We count thirty downloads for each user each month. No published plan holds 48 languages.
5 Lokalise publishes $144, $375 and $999 a month, and no key limit. We assume the three plans hold 5,000, 20,000 and 100,000 keys, and twelve words for each key. Human translation is billed apart.
6 Phrase publishes $525 a month for 1 million managed words on the software plan, and $1,245 for 3 million. We count one managed word for each word in each language, and we repeat the $1,245 block above 3 million.
7 Smartling does not publish a platform price. We count only the AI translation, at the published $0.06 a word, spread over twelve months. Your real bill is higher.
8 Tolgee bills keys, not words, so we count twelve words for each key. It publishes €49 for 2,000 keys, €179 for 5,000 and €499 for 20,000, billed annually. Above that we repeat the €499 block. Machine translation credits are counted apart.
9 Weglot translates a web page, not an app. Its largest published plan is €699 a month for 20 languages, so we scale that price to 48 languages. Your words fit inside every one of these plans.
Every other line grows with your product. Ours is a flat price.
Frequently asked questions
Do I need to manage translation files?
No. There are no JSON files, no key extraction step, and no build-time plugin. You write your sentences in your source language, the SDK sends them to the API on the first request, the API translates them with AI, and every subsequent request is served from the cache. Your code is the single source of truth.
What happens when I change the source text?
The source text is the key. Change it and you get a new key with a fresh AI translation. The old key simply stops being requested, and the dashboard cleans it up after 30 days without use. A typo fix is the same: fix it in the code, deploy, and the new text is translated on the first request.
How much does it cost?
Hosted: one flat price per project per month, by monthly active users — €4 up to 1,000, €9 up to 5,000, €29 up to 50,000, €99 up to 500,000 — billed monthly, cancel any time. Self-hosted: free to install, one project free forever, €30 once for unlimited projects, no subscription and no usage cap. There is no per-seat, per-word, or per-string pricing on either.
Can I export my translations?
Yes. The dashboard has an "Export JSON" button that downloads every translation in your project as a single JSON file. The export includes all rows, not just the ones used in the last 30 days, so it is a full backup. There is no lock-in: your data is always one click away.
One maintainer: what happens if the project stops?
The SDKs are MIT and stay on npm, PyPI, RubyGems and the other registries. The server is public under the Elastic License 2.0 and the protocol between them is documented, so anyone can run it, fork it, or write a server against it. Your translations leave in one JSON export at any time. If the hosted service stopped: installed apps keep serving their cached translations, a fresh client falls back to the source language, and pointing the SDK at your own instance with API_URL restores everything. What you would lose is the AI translation of new strings on i18n-keyless.com — not your data and not your app.
Read the full answer, with codeWatch the Demo
Watch this short video to see i18n-keyless in action
Get API Key for One App
Your whole i18n implementation in 5 minutes. iykyk.