Pronunciation is how you remember what you learned

I built konid after spending six months translating work emails in French and retaining almost nothing. I'd get a translation, paste it in, move on. Same words came up two weeks later and I had to look them up again — the task got done but nothing stuck.

The missing piece wasn't more vocabulary. It was audio. Hearing the actual vowel length, the liaison, the stress pattern — that's what converts a string of characters into something your brain files as language rather than text. So konid plays pronunciation through your speakers directly, using node-edge-tts with no external API key.

But pronunciation alone isn't enough if the phrase is wrong for the situation. Literal translation tools return one answer with no indication of whether it sounds like a native speaker, a textbook, or someone who learned the language from a phrasebook. konid returns three options per query, ordered casual to formal, with the register explained and a cultural context note comparing the options. If you're trying to tell your partner something that sounds real in their first language — not technically correct, actually real — the difference between those three options matters.

It runs as an MCP server, so it lives inside the tools you're already in. Claude Code setup is one line:

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, it installs as an app via Developer mode using the endpoint https://konid.fly.dev/mcp.

13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and more.

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