Every work email you translate could teach you something — most don't

I was translating professional emails in French every day for six months and my French barely moved. The tool gave me one answer, I pasted it, done. No sense of whether it was stiff or natural, formal or weirdly casual, or what word choice revealed about register.

So I built konid: you give it something you want to say, it gives you three translations ordered casual to formal, with each register explained and the differences called out explicitly. Not just "here's the formal one" — the response tells you why one phrasing is warmer, why another is safer for a new client, what a native speaker would actually use.

For the French email problem, that means seeing "Cordialement" vs "Bien à vous" vs "Salutations distinguées" side by side, with a note that the last one starts to read as ironic in modern business contexts. One query, one small thing retained.

The audio pronunciation runs through speakers directly via node-edge-tts — no external API key. It supports 13+ languages: French, Spanish, Mandarin, Japanese, Korean, Arabic, German, Portuguese, Italian, Russian, Hindi, and more.

It runs as an MCP server, so if your daily work is already in Claude Code, Cursor, Windsurf, or VS Code Copilot, you can add it in one line:

claude mcp add konid-ai -- npx -y konid-ai

Also works as a ChatGPT app via Developer mode with the endpoint https://konid.fly.dev/mcp.

The idea was to make every translation a small lesson rather than a one-shot utility. After a month of using it for real work emails, I started developing actual intuitions about register — not because I studied, but because the comparison was there every time I needed it.

MIT licensed: github.com/robertnowell/konid-language-learning