The intelligent CLI toolkit that automates your internationalization workflow with AI-powered translations.
Stop context switching. Let AI handle your translations while you stay in your code editor.
- 🔍 Smart Detection: Automatically scans your codebase to find translation keys
- 🤖 AI Translation: Generates high-quality translations using OpenAI or Gemini models
- 🔄 Sync & Maintain: Keeps all your locales in perfect sync
- 🧹 Clean & Optimize: Removes unused translations and creates optimized bundles
- ⚡ CI/CD Ready: Perfect for automated workflows and deployment pipelines
- 🔌 MCP Integration: Connect with Cursor and other LLMs to add translation keys on the fly
- JSON-based i18n libraries (react-i18next, next-i18next, vue-i18n, etc.)
- Node.js 16+
- An OpenAI or Google Gemini API key
Traditional workflow: 40+ minutes and 13 context switches to add 10 keys across 4 languages.
With i18n-magic + AI agents: 2.5 minutes, zero context switches. Stay in flow.
npx @scoutello/i18n-magic scan # Find & add missing translations
npx @scoutello/i18n-magic sync # Translate to all languages
npx @scoutello/i18n-magic replace [key] # Update existing translation
npx @scoutello/i18n-magic clean # Remove unused keys
npx @scoutello/i18n-magic check-missing # CI/CD validationscan - Scans your codebase for missing keys, prompts you for values, auto-translates to all locales.
sync - Takes your default locale translations and translates missing keys to other locales. Perfect for CI/CD.
replace - Update an existing translation key across all locales. Detects which namespaces use the key automatically.
clean - Removes unused translation keys from all locales. Great for keeping files lean.
check-missing - Dry-run check. Exits with error code if translations are missing. Perfect for CI pipelines.
// i18n-magic.js
globPatterns: [
'./src/shared/**/*.tsx', // → common.json
{ pattern: './src/dashboard/**/*.tsx', namespaces: ['dashboard'] },
{ pattern: './src/mobile/**/*.tsx', namespaces: ['mobile'] }
]Result: Separate files per feature (common.json, dashboard.json, mobile.json) across all locales.
MCP (Model Context Protocol) = API for AI agents.
Install the i18n-magic MCP server in Cursor, and your AI gets 5 new tools:
Fuzzy search across all translations. AI searches before adding anything.
Adds key to English (en) locale. Run sync afterward to translate to other languages.
Retrieve current value for any key.
Update a key and instantly translate to all languages. No sync needed!
Show all missing keys across your codebase.
npm install @scoutello/i18n-magicmodule.exports = {
globPatterns: ['./src/**/*.{ts,tsx,js,jsx}'],
loadPath: 'locales/{{lng}}/{{ns}}.json',
savePath: 'locales/{{lng}}/{{ns}}.json',
locales: ['en', 'de', 'es', 'fr'],
defaultLocale: 'en',
namespaces: ['common'],
defaultNamespace: 'common',
context: 'Your app description here',
model: 'gpt-4o-mini', // or 'gemini-2.0-flash-lite'
// Optional: Auto-clean before scanning
autoClear: true,
// Optional: Skip translation during scan (translate later with sync)
disableTranslationDuringScan: false,
}Key options:
context: Describe your app/domain. Better context = better translations.autoClear: Auto-remove unused keys before scanning.disableTranslationDuringScan: Settrueto only add English during scan, then usesynccommand for translations.
OPENAI_API_KEY=sk-your-key-hereSettings → Features → Model Context Protocol:
{
"i18n-magic": {
"command": "node",
"args": ["./node_modules/@scoutello/i18n-magic/dist/mcp-server.js"]
}
}Done! Test by asking: "Search for translations with 'password'"
You: "Add a submit button"
AI:
- Searches existing translations for "submit"
- Finds
submitButton: "Submit"already exists - Suggests reusing it
- Generates:
<Button>{t('submitButton')}</Button>
Result: No duplicate keys, instant code.
If nothing exists:
- AI adds key to English:
add_translation_key - You run:
npx @scoutello/i18n-magic sync - Translated to all languages!
You: "Change welcome message to 'Welcome back!'"
AI:
- Finds the key via search
- Calls
update_translation_key - Auto-translates to ALL languages instantly
No sync needed!
You: "What translations are missing?"
AI: Calls list_untranslated_keys, shows you the list.
Add them now or during development. Your choice.
Building a UserProfile component:
You: "Replace 'User Profile' with translation"
AI:
- Searches for "user profile"
- Nothing found
- Adds:
profile.title: "User Profile"to en/dashboard.json - Updates code:
<h1>{t('profile.title')}</h1>
You: "Add translations for labels"
AI:
- Searches "email" → finds
emailLabelin common - Reuses existing key (no duplicate!)
- Searches "full name" → nothing found
- Adds new:
profile.fullNameLabel: "Full Name"
You run:
npx @scoutello/i18n-magic syncResult: All keys now in EN, DE, ES, FR. Took 2 minutes.
Later: "Change title to 'Your Profile'"
AI: Calls update_translation_key → instantly updated in all 4 languages. No sync needed!
// Generic → Okay
context: 'Web application'
// Specific → Much better!
context: 'E-commerce for outdoor gear. Friendly tone. Target: adventurers.'More context = better AI translations.
# .github/workflows/auto-translate.yml
on:
pull_request:
paths: ['locales/en/**/*.json']
jobs:
translate:
steps:
- run: npx @scoutello/i18n-magic sync
- run: git commit -am "chore: auto-translate"English changes → auto-translated → committed. Zero manual work.
- <100 keys: Single
common.json - 100-500 keys: Split by feature (
auth,dashboard) - 500+ keys: Feature-based with auto-assignment (see setup section)
Store translations anywhere (S3, databases, CDNs):
// Example: S3 storage
const { S3Client, GetObjectCommand, PutObjectCommand } = require('@aws-sdk/client-s3')
const s3Client = new S3Client({
region: process.env.S3_REGION,
credentials: {
accessKeyId: process.env.S3_ACCESS_KEY,
secretAccessKey: process.env.S3_SECRET_KEY,
}
})
module.exports = {
loadPath: async (locale, namespace) => {
const response = await s3Client.send(
new GetObjectCommand({
Bucket: 'my-translations-bucket',
Key: `locales/${locale}/${namespace}.json`,
})
)
const data = await response.Body.transformToString()
return JSON.parse(data)
},
savePath: async (locale, namespace, data) => {
await s3Client.send(
new PutObjectCommand({
Bucket: 'my-translations-bucket',
Key: `locales/${locale}/${namespace}.json`,
Body: JSON.stringify(data, null, 2),
ContentType: 'application/json',
})
)
},
// ... other config
}Error: MCP error -32000: Connection closed
Fix: Use auto-detection (no cwd needed):
{
"i18n-magic": {
"command": "node",
"args": ["./node_modules/@scoutello/i18n-magic/dist/mcp-server.js"]
}
}Still broken? Check i18n-magic.js exists in project root:
ls -la i18n-magic.jsProblem: Keys added but not translated.
Fix: Run sync!
npx @scoutello/i18n-magic syncadd_translation_key only adds English. sync translates to other languages.
- Restart Cursor after adding MCP config
- Check Settings → Features → MCP shows "i18n-magic" as Connected
- Test: Ask AI to "search translations for 'password'"
npm install @scoutello/i18n-magic- Create
i18n-magic.jsconfig (see setup section) - Add
OPENAI_API_KEYto.env - Configure MCP in Cursor
- Start coding. AI handles translations.
5 minutes to set up. Hours saved every week.
- MCP Details: MCP-SERVER.md
- Example App: example-app/
- GitHub: github.com/BjoernRave/i18n-magic
- NPM: @scoutello/i18n-magic
Never context switch for translations again. 🌍