Every day I translate work emails in a language I'm still learning. For months I used standard translation tools — paste, copy, send. Fast, accurate enough, completely useless for actually learning the language. One answer, no register, no explanation of why that phrasing and not another.
The problem isn't accuracy. It's that a single translation gives you nothing to hold onto. You don't know if you just wrote something stiff and bureaucratic to a colleague who expects warmth, or casual to a client who expected formality. You copy-paste and move on, having learned exactly nothing.
So I built konid as an MCP server. For any phrase you want to say, it returns three options ordered casual to formal, with the register of each option explained and a nuance comparison between them. Not "here is the translation" — more like "here are the three ways a native speaker might say this, here's when each fits, here's what's different."
Pronunciation ships with it too: audio plays directly through your speakers via node-edge-tts, no external API key. The 13+ supported languages include Mandarin, Japanese, Korean, Spanish, French, and Arabic, among others.
The install is one line in Claude Code:
claude mcp add konid-ai -- npx -y konid-ai
Also works in Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. If you're on ChatGPT with Developer mode, add the endpoint https://konid.fly.dev/mcp as a custom tool.
The daily translation task doesn't disappear — the emails still need sending. But now each one leaves something behind: a register choice I made consciously, a nuance I noticed, a phrase I'll recognize next time I hear it.
Top comments (0)