I wanted to tell my girlfriend 'I missed you today' in Farsi. Google Translate gave me one answer. It was technically correct the way a dictionary entry is correct β no sense of how it would land, whether it was something a person actually says, or how formal it sounded. I said it anyway. She laughed, not unkindly.
That's the gap literal translation tools don't close. They get you through the task. They don't teach you anything, and they don't tell you when you've chosen the wrong register.
So I built konid β a language coach that returns three options per query, ordered casual to formal, with cultural context and nuance notes for each. The register is explained, not assumed.
The audio piece matters more than I expected. Playing pronunciation through speakers (via node-edge-tts, no external API key required) isn't about confirming you spelled something right. It's about tone contour and vowel length β the things that change whether a phrase sounds like a native speaker said it or like someone sounding it out. In tonal languages like Mandarin, hearing the difference between a third-tone and a fourth-tone on 'mai' is information no romanization system fully conveys. In Japanese, vowel length changes meaning entirely. Reading the text, you can fool yourself into thinking you know how something sounds. Hearing it, you find out.
The three-option structure does something the audio can't do alone: it shows you the space around what you want to say. 'I missed you' in Spanish has a version you'd text a close friend, a version you'd say to someone you're still getting to know, and a version that lands differently depending on whether the relationship is romantic or not. Returning all three, with the comparison explained, turns each query into a short lesson rather than a lookup.
konid runs as an MCP server β claude mcp add konid-ai -- npx -y konid-ai drops it into Claude Code, and it also works with Cursor, VS Code Copilot, Windsurf, Zed, JetBrains, and Claude Cowork. If you're using ChatGPT, there's a Developer mode endpoint at https://konid.fly.dev/mcp. Covers 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and others.
MIT licensed.
Top comments (0)