I was translating work emails in Japanese every day — copy-pasting into Google Translate, getting one answer back, sending it, and learning nothing. A year in, I still couldn't tell you whether the phrase I'd used forty times was casual, formal, or accidentally rude. The literal translation worked in the sense that nobody complained. But it wasn't teaching me anything about how the language actually moves between registers.
konid is an MCP server I built to fix that. For any phrase you want to say, it returns three options ordered casual to formal, each with the register explained and cultural context for why you'd choose one over another. Then it plays audio pronunciation through your speakers directly — no external API key, using node-edge-tts under the hood.
The register gap is what makes or breaks real communication. In Japanese, the difference between casual and formal isn't a footnote — it's the whole social signal. Same in Korean. Even in Spanish, telling someone "te extrañé" versus "lo extrañé" versus something more tender isn't about grammar, it's about what you're telling them about the relationship. A tool that hands you one answer collapses all of that into noise.
Here's what using it looks like in Claude Code:
claude mcp add konid-ai -- npx -y konid-ai
After that, you can ask it anything in natural language — "how do I say I missed you today in Japanese, and what's the difference between the options" — and get back three choices with nuance notes and audio. It works the same way in Cursor, Windsurf, VS Code Copilot, Zed, JetBrains, and Claude Cowork. If you're on ChatGPT, there's also a Developer mode endpoint at https://konid.fly.dev/mcp.
It covers 13+ languages: Mandarin, Japanese, Korean, Spanish, French, German, Portuguese, Italian, Russian, Arabic, Hindi, and more.
The name, for what it's worth: konid (کنید) is Farsi for "do" — take action.
MIT licensed. https://github.com/robertnowell/konid-language-learning
Top comments (0)