I translate work emails in a language I'm still learning. For two years, I used Google Translate or DeepL: paste text, copy result, send email, learn nothing. Each translation was a one-shot utility. I got through the task and stayed exactly as fluent as before.
The problem isn't accuracy. Those tools translate accurately. The problem is they give you one answer with no register, no sense of whether you're being too stiff or too casual, no explanation of why one phrasing works and another doesn't. You're copying output, not building any model of the language.
I built konid to close that loop. It returns three options for anything you want to say, ordered casual to formal, with the register explained and cultural nuance notes comparing the options. The point is that you see why the formal version is formal — what word choice or grammatical structure signals deference — not just that it exists.
Install as an MCP server in Claude Code:
claude mcp add konid-ai -- npx -y konid-ai
Works in Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. Also installs as a ChatGPT app via Developer mode with the endpoint https://konid.fly.dev/mcp.
For audio: it plays pronunciation through your speakers directly using node-edge-tts — no external API key. Hearing the register difference between a casual and formal option in Mandarin or Japanese is more useful than reading a note that says "this one is more polite."
Supports 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and others.
The daily-email use case is the one I built it for. Translating a work email and seeing that your instinct was the most formal register when a colleague would say something warmer — that's the feedback loop that sticks.
Top comments (0)