I wanted to tell my girlfriend 'I missed you today' in Farsi. Google Translate gave me one answer. It was technically correct and completely wrong for the moment — the kind of sentence you'd read off a form, not say to someone you love.
That gap between technically correct and actually right is where most translation tools stop caring. They hand you a string and move on. No indication of register. No sense of whether you're being formal with a stranger or warm with a partner. No way to know if you've chosen the phrase that fits.
So I built konid — an MCP server that returns three options per query, ordered casual to formal, with the register explained and a nuance comparison between them. Audio pronunciation plays directly through your speakers via node-edge-tts, no external API key.
The three-option structure matters more than it sounds. In Japanese, the distance between casual and polite speech isn't vocabulary — it's verb endings and entire grammatical registers that signal relationship, hierarchy, and context. Returning one answer flattens all of that. Returning three, with the distinctions explained, forces you to make a conscious choice. That choice is the actual language skill.
For translating work emails in a language you're still learning, it's the same problem in reverse: you want to get through the task and understand what you just sent. konid gives you the cultural context alongside the options — why one phrasing lands as professional and another lands as distant.
Installs in Claude Code with one command:
claude mcp add konid-ai -- npx -y konid-ai
Also works with Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. ChatGPT users can add it via Developer mode with the endpoint https://konid.fly.dev/mcp.
Supports 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and more.
MIT licensed. https://github.com/robertnowell/konid-language-learning
Top comments (0)