convez.ai
MCP Server
lopatnov

translate-mcp

lopatnov/translate-mcp

Ver no GitHub
BomPluginScore · de 0 a 100

MCP server for self-hosted speech and text translation — wraps Lopatnov.Translate gRPC backend and exposes transcription, translation, and language detection as AI tools for Claude and other MCP clients.

Instalação

$npx convez-ai lopatnov/translate-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
85%
Último commit
há 3 meses
Versão
v.3.0.0
Issues abertas
5
Contribuidores
4

README

A tradução para o português deste README ainda está na fila.

@lopatnov/translate-mcp

npm NPM version License GitHub stars

MCP server for Lopatnov.Translate — brings self-hosted speech-to-text, text translation, and language detection into Claude and any MCP-compatible AI client. All inference runs locally — no cloud, no API keys, no data leaves your machine.

Table of Contents

Prerequisites

  1. Node.js 20+

  2. Lopatnov.Translate gRPC service running (default: localhost:5100)

    Start with Docker:

    git clone https://github.com/lopatnov/translate.git
    cd translate
    # Download models — see docs/models.md
    docker compose -f docker/docker-compose.yml up
    

Installation

npm install -g @lopatnov/translate-mcp

Or run without installing:

npx -y @lopatnov/translate-mcp

AI Client Configuration

Claude Desktop

Add to claude_desktop_config.json:

  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
{
  "mcpServers": {
    "translate": {
      "command": "npx",
      "args": ["-y", "@lopatnov/translate-mcp"],
      "env": {
        "TRANSLATE_GRPC_URL": "localhost:5100"
      }
    }
  }
}

Claude Code

claude mcp add translate -e TRANSLATE_GRPC_URL=localhost:5100 -- npx -y @lopatnov/translate-mcp

Manage the server:

claude mcp list
claude mcp get translate
claude mcp remove translate

Cursor

Add to ~/.cursor/mcp.json (global) or .cursor/mcp.json (project):

{
  "mcpServers": {
    "translate": {
      "command": "npx",
      "args": ["-y", "@lopatnov/translate-mcp"],
      "env": {
        "TRANSLATE_GRPC_URL": "localhost:5100"
      }
    }
  }
}

Or via Cursor Settings → MCP → Add Server.

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "translate": {
      "command": "npx",
      "args": ["-y", "@lopatnov/translate-mcp"],
      "env": {
        "TRANSLATE_GRPC_URL": "localhost:5100"
      }
    }
  }
}

Or via Windsurf Settings → Cascade → MCP Servers → Add.

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in your workspace:

{
  "servers": {
    "translate": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@lopatnov/translate-mcp"],
      "env": {
        "TRANSLATE_GRPC_URL": "localhost:5100"
      }
    }
  }
}

Or via Command Palette → MCP: Add Server.

Zed

Add to Zed settings.json (Zed → Settings → Open Settings):

{
  "context_servers": {
    "translate": {
      "command": {
        "path": "npx",
        "args": ["-y", "@lopatnov/translate-mcp"]
      },
      "settings": {
        "TRANSLATE_GRPC_URL": "localhost:5100"
      }
    }
  }
}

Environment Variables

VariableDefaultDescription
TRANSLATE_GRPC_URLlocalhost:5100Host and port of the Lopatnov.Translate gRPC service

Tools

ToolDescription
translate_textTranslate a string between any two supported languages
detect_languageDetect the language of a text string
translate_localizationTranslate all strings in a JSON i18n file, preserving key structure
transcribe_audioTranscribe a WAV audio file to text (requires Whisper model)
get_capabilitiesList available translation models and service status

translate_text

ParameterTypeRequiredDescription
textstring✅Text to translate
target_languagestring✅Target language code, e.g. "uk", "de"
source_languagestring—Source language, or "auto" for auto-detection (default)
modelstring—Model name from config, e.g. "m2m100_418M". Omit for default
language_formatstring—"bcp47" (default) or "native" (model-specific codes)

detect_language

ParameterTypeRequiredDescription
textstring✅Text to detect
language_formatstring—Format of the returned code. Default: "bcp47"

translate_localization

ParameterTypeRequiredDescription
jsonstring✅Source JSON as a string
source_languagestring✅Source language (BCP-47)
target_languagestring✅Target language (BCP-47)
modelstring—Model name. Omit for default
existing_translationstring—Previously translated JSON — matching keys are reused
language_formatstring—Language code format. Default: "bcp47"

transcribe_audio

ParameterTypeRequiredDescription
file_pathstring✅Absolute path to the WAV file
languagestring—Language hint (BCP-47), or "auto" for Whisper auto-detection
language_formatstring—Format of the returned detected language. Default: "bcp47"

get_capabilities

Returns available translation models and whether STT is enabled. No parameters.


Contributing

Contributions are welcome! Please read CONTRIBUTING.md before opening a pull request.


Built With


License

Apache-2.0 © 2026 Oleksandr Lopatnov · LinkedIn

Ver o original no GitHub

Bom para

Relacionados