I was drafting a message to my girlfriend in Farsi — something small, just that I'd missed her that day — and Google Translate gave me one option. No indication of whether it was tender or clinical, whether a native speaker would find it warm or weirdly formal. I sent it anyway and she laughed, kindly. It sounded like a receipt.
So I built konid.
The core behavior: you describe what you want to say, and it returns three translations ordered casual to formal, with a note explaining the register difference between them. Not "here's the word" but "here's what each one signals to a native speaker and when you'd reach for it."
For the Farsi version of "I missed you today" the three options ranged from something you'd text a close partner to something that reads as almost literary — and the nuance note explained which one actually sounds like a person rather than a phrase card. Audio pronunciation plays through your speakers via node-edge-tts, no API key needed.
It covers 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, Farsi, and more.
The setup I use daily:
claude mcp add konid-ai -- npx -y konid-ai
That installs it as an MCP server in Claude Code. It also works in Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. If you're on ChatGPT, there's a Developer mode install via https://konid.fly.dev/mcp.
I kept reaching for it when drafting work emails in Spanish — not just to get through the task but to actually understand why one phrasing felt more professional than another. The register comparison does that work.
MIT licensed. github.com/robertnowell/konid-language-learning
Top comments (0)