Polyglot MCP translate everything, locally.

Private

Translations run on your GPU. No text leaves your machine unless you point Polyglot at a remote Ollama yourself.

Fast

About a second per translation once the default 27B model is loaded, and ~600ms with 12B on a 16 GB GPU.

Free

No API keys, no subscriptions, no rate limits. Ollama + TranslateGemma are fully open source (Apache 2.0).

57 Languages

From Afrikaans to Vietnamese — including CJK, Arabic, Hindi, and all major European languages.

Smart Chunking

Long text is split at paragraph and sentence boundaries, preserving context across chunks.

Model Choice

Defaults to 27B for maximum accuracy; switch to 12B or 4B for smaller GPUs or more speed.

Claude Code / Claude Desktop

// Add to your MCP config (.mcp.json)
{
  "mcpServers": {
    "polyglot": {
      "command": "npx",
      "args": ["-y", "@mcptoolshop/polyglot-mcp"]
    }
  }
}

Prerequisites

# 1. Install Ollama (ollama.com)
# 2. Pull the translation model:
ollama pull translategemma:27b  # 17 GB (or :12b, 8.1 GB, on smaller GPUs)

# 3. Start Ollama:
ollama serve