I was drafting a message to my girlfriend in her first language. Google Translate returned one option. No indication of whether it was the right register, whether it sounded stiff or natural, whether a native speaker would wince reading it.
That's the actual failure mode of literal translation tools: they collapse the choice space to a single answer and hand it to you with no context. If you're still learning a language, you can't tell what you lost.
So I built konid — an MCP server that returns three options per query, ordered casual to formal, with the register explained and cultural context comparing the options. Audio pronunciation plays through your speakers directly, no API key required (it uses node-edge-tts under the hood).
The install for Claude Code is one line:
claude mcp add konid-ai -- npx -y konid-ai
It also works with Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. If you're on ChatGPT, you can add it via Developer mode using https://konid.fly.dev/mcp as the endpoint.
For something like "I missed you today" in Japanese, you don't want a single answer. You want to know that 会いたかった lands differently than 今日はあなたのことが恋しかったです — and why. That's what the three options with nuance notes give you: actual signal about what you're choosing between.
Supports 13+ languages including Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, and Hindi.
MIT licensed. github.com/robertnowell/konid-language-learning
Top comments (0)