AI-powered localization — translate, manage translation memory, and access style guides.
Runs on your machine from its package; the steps are below.
At most: creates or changes data: records, files, settings.
Odel charges nothing per call. The publisher may charge for its own service.
Runs locally over stdio
This server isn't hosted — your MCP client launches it from a package registry. Use one of the commands below, or drop the config into your client (e.g. Claude Desktop).
Run
uvx nativ-mcp
MCP client config
{
"mcpServers": {
"nativ": {
"command": "uvx",
"args": [
"nativ-mcp"
]
}
}
}mcp-name: io.github.Nativ-Technologies/nativ
AI-powered localization for any MCP-compatible tool — Claude Code, Cursor, Windsurf, and more.
Nativ is a localization platform that uses AI to translate content while respecting your brand voice, translation memory, glossaries, and style guides. This MCP server brings Nativ's full localization engine into your AI coding workflow.
Sign up at dashboard.usenativ.com, go to Settings → API Keys, and create a key. It looks like nativ_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx.
Add to your MCP configuration:
~/.claude/claude_desktop_config.json){
"mcpServers": {
"nativ": {
"command": "npx",
"args": ["-y", "nativ-mcp"],
"env": {
"NATIV_API_KEY": "nativ_your_api_key_here"
}
}
}
}
.cursor/mcp.json in your project or ~/.cursor/mcp.json globally){
"mcpServers": {
"nativ": {
"command": "npx",
"args": ["-y", "nativ-mcp"],
"env": {
"NATIV_API_KEY": "nativ_your_api_key_here"
}
}
}
}
{
"mcpServers": {
"nativ": {
"command": "npx",
"args": ["-y", "nativ-mcp"],
"env": {
"NATIV_API_KEY": "nativ_your_api_key_here"
}
}
}
}
Note:
npxauto-downloads the package on first run — no manual install needed. Ifuvisn't already on your machine, it will be installed automatically on first launch.Alternative: use
uvxdirectlyIf you already have
uvinstalled and prefer to skip the npm wrapper:json{ "mcpServers": { "nativ": { "command": "uvx", "args": ["nativ-mcp"], "env": { "NATIV_API_KEY": "nativ_your_api_key_here" } } } }macOS tip: If you get
spawn uvx ENOENTin Cursor or Claude Desktop, GUI apps don't inherit your shell PATH. Use the full path (e.g."command": "/Users/you/.local/bin/uvx") or wrap in a login shell:"command": "/bin/sh", "args": ["-lc", "uvx nativ-mcp"].
Ask your AI assistant things like:
| Tool | Description |
|---|---|
translate | Translate text using the full localization engine (TM, style guides, brand voice, glossary) |
translate_batch | Translate multiple texts to a target language in one call |
search_translation_memory | Fuzzy-search the translation memory for existing translations |
add_translation_memory_entry | Add an approved translation to TM for future reuse |
get_languages | List all configured languages with formality and style settings |
get_translation_memory_stats | Get TM statistics — total entries, sources, and breakdown |
get_style_guides | List all style guides with their content and status |
get_brand_voice | Get the brand voice prompt that shapes all translations |
extract_image_text | Extract on-image text segments (OCR) in reading order |
localize_image | Replace on-image text with translations in place, keeping layout and graphics intact |
check_image_cultural_fit | Flag cultural-sensitivity issues in an image across target markets |
transcribe_audio | Speech-to-text for a local audio or video file (draft; billed on synthesize) |
parse_audio_script | Parse a script (.txt/.docx/.pdf/.srt) instead of running STT |
list_voices | List TTS voices for dubbing |
preview_voice | Save a short TTS preview WAV to disk |
get_voice_consent_script | Consent copy required for voice cloning |
synthesize_audio | Generate billed localized TTS (and optional dubbed MP4) and save to disk |
register_video | Store a source video for later remux |
remux_video | Mux a WAV onto a registered video |
parse_subtitle | Parse a local SRT/ITT into timed cues |
localize_subtitles | Localize subtitles in-memory from SRT/ITT and/or video STT |
| URI | Description |
|---|---|
nativ://languages | Configured languages (JSON) |
nativ://style-guides | All style guides (JSON) |
nativ://brand-prompt | Brand voice prompt (JSON) |
nativ://tm/stats | Translation memory statistics (JSON) |
| Prompt | Description |
|---|---|
localize-content | Guided workflow to localize content into target languages |
review-translation | Review a translation against TM, style guides, and brand voice |
batch-localize-strings | Batch-localize i18n strings with structured output |
You: Translate "The future of luxury, delivered" to French and Japanese
AI: [calls translate tool for each language]
Translation (French): "L'avenir du luxe, livré chez vous"
TM Match: 0% — new translation, no prior TM entries
Rationale: "Livré chez vous" adds a personal touch absent from the literal
"livré", aligning with the brand's premium yet approachable voice.
Translation (Japanese): "ラグジュアリーの未来を、あなたの元へ"
TM Match: 45% partial — similar pattern found in TM from brand_voice source
You: Do we have translations for "Add to cart" in our TM?
AI: [calls search_translation_memory]
TM Search Results for "Add to cart" (3 matches):
- 95% [strong] "Add to cart" → "Ajouter au panier" (source: approved)
- 95% [strong] "Add to cart" → "In den Warenkorb" (source: brand_voice)
- 72% [partial] "Add items to cart" → "Ajouter des articles" (source: phrase_tm)
You: Localize these to French:
- "Sign up"
- "Log in"
- "Forgot password?"
- "Continue with Google"
AI: [calls translate_batch]
Batch translation to French (4 items):
1. "Sign up" → "S'inscrire" (TM 100%)
2. "Log in" → "Se connecter" (TM 100%)
3. "Forgot password?" → "Mot de passe oublié ?" (TM 92%)
4. "Continue with Google" → "Continuer avec Google" (TM 85%)
| Environment Variable | Required | Description |
|---|---|---|
NATIV_API_KEY | Yes | Your Nativ API key (nativ_xxx...) |
NATIV_API_URL | No | API base URL (defaults to https://api.usenativ.com) |
This MCP server acts as a bridge between your AI coding assistant and the Nativ API:
┌─────────────────────┐ ┌──────────────┐ ┌─────────────────┐
│ Claude / Cursor / │────▶│ Nativ MCP │────▶│ Nativ API │
│ Windsurf / etc. │◀────│ Server │◀────│ (Translation, │
│ │ │ (stdio) │ │ TM, Styles) │
└─────────────────────┘ └──────────────┘ └─────────────────┘
The MCP server runs locally via stdio. It authenticates with your API key and calls the Nativ REST API on your behalf. Your AI assistant sees Nativ's tools, resources, and prompts as native capabilities.
# Clone the repo
git clone https://github.com/nativ-ai/nativ-mcp.git
cd nativ-mcp
# Set up environment
uv venv
source .venv/bin/activate
uv pip install -e .
# Run the server (for testing)
NATIV_API_KEY=nativ_xxx nativ-mcp
# Run with MCP Inspector
NATIV_API_KEY=nativ_xxx npx @modelcontextprotocol/inspector uv run nativ-mcp
MIT — see LICENSE.