I was learning to write work emails in French. I could get through them with Google Translate, but I kept making the same register mistakes — using informal constructions with people I'd never met, or stiff phrasebook phrases with colleagues I'd worked with for months. The problem wasn't vocabulary. It was that I had no mental model of what formal-versus-casual actually sounds like in French, so I couldn't calibrate.
konid is an MCP server I built that returns three translation options per query, ordered casual to formal, with the register explained for each. But the part that changed how I actually retain things is the audio: it plays pronunciation directly through your speakers using node-edge-tts, no external API key needed.
Hearing the tone contour and vowel length of all three options back-to-back does something that reading them doesn't. In Mandarin especially, the difference between registers involves phonetic changes that are invisible in pinyin. In Japanese, the formal and casual forms of the same sentence have different rhythm. When you hear the contrast instead of reading it, you build a sound-memory, not just a character-memory — and sound-memory is what you actually use when you're speaking or listening in real time.
Install in Claude Code:
claude mcp add konid-ai -- npx -y konid-ai
Or in Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, or Claude Cowork via the MCP server config. It also runs as a ChatGPT app in Developer mode at https://konid.fly.dev/mcp.
Supports 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and others. MIT licensed.
Top comments (0)